The incidence of leukaemia and non-Hodgkin’s lymphoma in young people living near certain nuclear establishments in the United Kingdom has been the subject of much research. ^1–23 Of particular concern has been the sustained increased incidence of these cancers in children and young adults living in the Cumbrian village of Seascale, near British Nuclear Fuels’ Sellafield reprocessing plant.^1,6,7,18,21 In addition, increased rates of leukaemia and non-Hodgkin’s lymphoma have been reported in young people living near the Atomic Energy Authority’s Dounreay plant^{4 9}and the Atomic Weapons Establishments at Aldermaston and Burghfield.^{8 13}

The nuclear industry family study was set up to investigate possible links between child health and parents’ occupational exposure to ionising radiation.²⁴ Gardner et al’s report,¹⁰ suggesting an association between leukaemia in the under 25s and paternal preconceptual exposure to external sources of ionising radiation at work, was published while our study was still at the planning stage. The interpretation of Gardner et al’s findings^10,12 continues to be debated,^19,25–30 and this paper focuses on our findings for cancer in the under 25s.

Information about the study is given elsewhere.²⁴ Briefly, the study population comprised employees of three nuclear authorities: the Atomic Weapons Establishment, the Atomic Energy Authority, and British Nuclear Fuels. All current employees of these three authorities, and past employees of the last two under 75 years old whose details were recorded on the pensions database, were surveyed over the four years 1993 to 1996. For the purposes of the study, employers provided each subject with a unique personal identifier, which was used to link respondents’ data to industry employment and monitoring records (including annual radiation dose).

After undelivered post was excluded, the response rate was 82% for male workers and 88% for female workers, with 39557 live births being reported by 18131 fathers and 8883 by 4435 mothers.²⁴ The total number of children in the study was 46107; 2333 children had both a mother and father in the surveyed population. When a parent reported that one of their children had had a serious illness, such as cancer, more information was requested about the condition and signed consent was sought to access relevant medical notes. Parents were asked to forward consent forms to living children aged 18 years or over.

External comparisons We obtained annual cancer incidence rates specific for sex and five year age groups for England and Wales, 1971-89 from the Office for National Statistics. National cancer registration in England and Wales did not have an acceptable level of coverage until 1971,³²and data for the 1990s were unavailable at the time of analysis. We therefore decided in advance to restrict the comparison with the general population to births occurring in 1965 or later, using average national rates for the five years 1971-5 to estimate rates before 1971 and average rates for the five years 1985-9 to estimate rates after 1989. For births occurring in 1965 or later, offspring-years at risk were stratified by sex, age (five year groups), and calendar period (single years), and the numbers of cancers expected on the basis of the England and Wales rates were calculated. Exact 95% confidence intervals and P values for the resulting standardised incidence ratios were based on the Poisson distribution.³³

Internal cohort comparisons We estimated the effect of parental exposure to radiation on the risk of malignancy by rate ratios (hazard ratios) using Cox proportional hazards modelling, with age as the time variable and adjusting for calendar period in 10 year intervals and child’s sex.³⁴ In analyses of radiation exposure at any time, parental exposure wastreated as a time dependent variable. Hence, the same child could contribute offspring-years both to the “unexposed” category (offspring-years occurring before first parental exposure) and to one or more of the “exposed” categories (offspring-years occurring after first parental exposure). For analyses relating to parental exposures occurring before conception, however, all offspring-years were assigned to the same exposure group. Trends of increasing risk with cumulative preconceptual radiation dose were tested by using Cox proportional hazards modelling. Only children whose fathers had been monitored for external sources of ionising radiation before their conception were included in such analyses, the actual preconceptual dose accumulated being modelled using linear, quadratic, or higher order terms.

Because many workers had more than one child in the study, standard errors for the rate ratios were not calculated by standard methods, which assume independence between observations. Instead, we used a robust method based on the “sandwich estimate” of the standard error in all regression analyses.^35,36 We then assessed the significance of the contribution of variables to the model using the Wald test with the robust variance estimate.^{37 38}

The hypothesis that occupational preconceptual exposure of fathers to external sources of ionising radiation increases the risk of leukaemia in their offspring derives mainly from the case-control study conducted by Gardner and colleagues.¹⁰ Gardner et al studied leukaemia and non-Hodgkin’s lymphoma in those born in west Cumbria and diagnosed there before their 25th birthday between 1 January 1950 and 31 December 1985.¹¹ Because of the potential for overlap with our study, analyses relating to paternal preconceptual exposure were conducted both with and without children born in Cumbria on, or before, 31 December 1985; this last group comprised 5950 (15%) children of male workers and 1170 (13%) of female workers.

The 39557 children of male workers and 8883 children of female workers respectively contributed 716325 and 156304 offspring-years at risk to the analysis. The median length of follow up (up to a maximum 25 years) was 23 years for children of male workers and 22.7 years for children of female workers (table 1). At the time of survey, the oldest child was 58 (born in 1937) and the youngest was less than a month (born in 1996). The proportion of parents monitored for exposure to ionising radiation at some time before their child’s 25th birthday was 80% for male employees compared with 25% for female employees. Likewise, although 40% of children of male employees had a father who was monitored before their conception, only 9% of children of female employees had a mother who was similarly exposed (table 1).

Table 1 Description of children and malignancies diagnosed before the age of 25 years, reported by parents employed by nuclear authorities. Values are numbers (percentages) of children unless stated otherwise

A total of 111 children were reported to have developed a malignancy before their 25th birthday. Five of these, none of whom had leukaemia, had both parents in the study. The earliest year of diagnosis was 1951 and the latest was 1993. Confirmatory evidence in the form of a cancer registration, death certificate, or entry in medical notes was located for 108 (97%) of the 111 children.

Table 2 Numbers of cancers diagnosed in children born in 1965 or later compared with number expected on basis of cancer incidence rates for England and Wales

Parental employment and monitoring at any time Table 3 gives the offspring-years at risk and cancers among children of nuclear industry workers classified according to their parents’ history of employment in the industries surveyed and monitored for radiation exposure at any time before their child’s 25th birthday. For each malignancy group the baseline for all comparisons is the cancer rate in offspring before their parent had a record of employment of monitoring. Data in the second category relate to cancer risk after parental first employment but before first monitoring, and data in the third category to cancer risk after first monitoring.

Table 3 Cancers and offspring-years at risk in children of nuclear workers, classified according to parental history of employment and monitoring for potential exposure to ionising radiation

Relatively few children had mothers who were monitored for external sources of ionising radiation (table 3). Indeed, the rarity of this occupational exposure in women, coupled with the rarity of cancer in the under 25s resulted in small numbers of cases in most exposure categories. For example, no leukaemias were reported in children of women who were monitored for exposure to external sources of ionising radiation. Because of lack of data, no further analyses of cancer in children of female employees are presented here.

Among children of male workers, there were no significant variations in cancer risk with paternal history of employment in the nuclear industry or with paternal monitoring for radiation exposure (table 3). Furthermore, there was no evidence that the pattern of rate ratios varied with employer, date of first employment, or whether the respondent was a current or past worker at the time of the survey (data not shown).

Paternal employment before conception Table 4 gives the findings relating to paternal employment, monitoring status, and estimated dose of ionising radiation in the preconceptual period. The preconceptual dose groupings are similar to those used by Gardner and colleagues,^10,12 except that children of fathers with a record of monitoring for exposure to ionising radiation for whom there was no recorded dose are included in the lowest dose group (in our data none of the fathers of children with malignancies had recorded doses below 0.1 mSv).

Table 4 Cancers under age of 25 years in children of male nuclear workers, classified according to paternal employment history before conception, monitoring for radiation exposure, and estimated radiation dose with separate analysis excluding children who may (more ...)

Table 4 includes more cancers and offspring-years in the baseline category (and correspondingly fewer in the employed and monitored categories) than table 3 becausepaternal monitoring and employment is considered only to the point of the child’s estimated date of conception. For leukaemia, the estimated rate ratio in children whose fathers had a record of monitoring for radiation exposure before their conception was 2.2 (0.8 to 6.1; based on 12 exposed cases). Among children whose fathers were monitored for external sources alone it was 2.1 (0.7 to 6.5) and among children whose fathers were also monitored for radionuclides it was 2.4 (0.7 to 8.3). With respect to recorded whole body dose, the rate ratio for leukaemia in those whose fathers had accumulated a lifetime dose of 100 mSv before their child’s conception was 5.8 (1.3 to 24.7). This excess was based on three exposed cases, and the fathers of these three children also had estimated doses in the six months before conception of 10 mSv or more (rate ratio 7.7; 1.9 to 31.0). However, among those who had been monitored for external radiation exposure before their child’s conception, there was no significant dose-response relation (P values for linear and quadratic terms being 0.18 and 0.16 for lifetime exposure and 0.24 and 0.28 for the six months before conception).

The rate ratios for leukaemia calculated after children born in Cumbria before 1986 (potential “Gardner” cases) were excluded are based on even smaller numbers, particularly in the high dose categories (table 4). The numbers of offspring-years at risk, however, are still substantial and the results for leukaemia are similar to those calculated for the total data. Nevertheless, it is important to note that the rate ratio estimates of 6.6 (0.7 to 67.1) for a total cumulative preconceptual dose 100 mSv and 11.0 (1.2 to 105.0) for a cumulative dose 10 mSv in the six months before conception were based on only one case.

The raised rate ratios for all malignancies and for leukaemia and non-Hodgkin’s lymphoma combined among all children whose fathers accrued a dose of 100 mSv before their conception (and 10 mSv in the six months before conception) were mainly due to leukaemia. The effect of leukaemia was less when Cumbrian births were excluded, the rate ratios for malignancies other than leukaemia and non-Hodgkin’s lymphoma being 4.0 (1.2 to 13.6, based on three exposed cases) for a lifetime paternal preconceptual dose 100 mSv and 4.4 (1.1 to 18.6, two exposed cases) for a dose 10 mSv in the six months before conception.

So that our results could be compared directly with the record linkage study,^22,23 the analyses were repeated for the first 15 years of life. These results, which are broadly similar to those for 0-24 year olds, are available on the BMJ’s website.

The Nuclear Industry Family Study differs in concept and design from other studies that have examined the relation between parental employment in the nuclear industry and child health. It is a within industry investigation, being more analogous to the cohort studies that examined the health of nuclear workers^39–44 than to the population based case-control studies that have reported on cancer in young people living in areas around nuclear sites.^{10,14,16,45,46} The study was resticted to employees of nuclear establishments operated by the Atomic Energy Authority, Atomic Weapons Establishment, and British Nuclear Fuels because each of these industries had at least one plant that had been subject to a report alleging an increased incidence of leukaemia in young people residing in its vicinity.^{1,4,6,9,10 14}

In common with all studies of cancer in offspring of nuclear workers, our findings are based on small numbers of cancers. Although in the case-control studies the small numbers are principally due to the comparative rarity of employment in the nuclear industry,^{10,14,16,22,45,46} even in areas close to the plants, in our study they largely reflect the rarity of cancer in young people. Of the 46107 children included, only 111 were reported to have had cancer diagnosed before their 25th birthday and only 28 had leukaemia.

The population studied by Gardner et al overlaps with ours, and the similarity between the findings must be interpreted with this in mind. We attempted to overcome this problem by conducting some analyses excluding children born in Cumbria before 1986, thereby excluding all children with the potential to have been in the west Cumbrian study. Monitored Sellafield workers had been exposed to higher doses of external ionising radiation, on average, than other workers in the study. Exclusion of children born in Cumbria before 1986 therefore removed proportionally more children from the high dose group than from the low dose group. For example, although all eight cases of leukaemia remained in the <50 mSv paternal preconceptual dose category, only one remained in the 100 mSv dose category (table 4). For leukaemia, the risk estimates for the high dose categories were similar to those obtained for the total data. These estimates were based, however, on a single case in the highest exposure categories. Furthermore, exclusion of Cumbrian births increased the risk estimates for malignancies other than leukaemia and non-Hodgkin’s lymphoma in the higher exposure categories.

Table 5 Summary of studies that have examined cancer risk of cancer in children whose fathers were monitored for occupational exposure to ionising radiation before their conception.* Values are risk ratios (95% confidence intervals) unless stated (more ...)

Table 6 Further details of studies examining cancer risk in children whose fathers were monitored for occupational exposure to ionising radiation

Our study has the potential for overlap with all but the Canadian study. For Caithness and west Berkshire, the potential for overlap is small for children whose fathers were monitored before their conception and negligible for those whose fathers had doses of 100 mSv or more. We removed cases from west Cumbria to avoid overlap with Gardner et al’s study, as did the record linkage study. Case overlap between our study and the record linkage study is harder to overcome because both are national investigations. At first sight the fact that both studies estimated an 80% increase in the risk of leukaemia and non-Hodgkin’s lymphoma among children of monitored workers might be attributed to overlapping data, as around half the fathers in the record linkage study were employed at some time by the industries studied here.²² However, in the record linkage study the significantly raised risk of 1.8 (1.1 to 3.0) for leukaemia and non-Hodgkin’s lymphoma had its origins in an estimated eightfold increase in risk (95% confidence interval 1.2 to ∞) among children with fathers whose estimated doses were below 0.1 mSv.^22,23 The removal of potential Gardner cases had no effect on this estimate. This observation, coupled with the knowledge that such low doses were rare in the industries studied here,²⁴ suggests that the excess risk in the record linkage study has its origins elsewhere. The reasons for the disagreement between the studies with data on children whose fathers were exposed to 100 mSv before their conception are unclear.

Although our study is comprehensive, cancer in young people is rare and our findings are based on small numbers of cases. Our data are too few to break down into non-overlapping exposure periods; fathers exposed to high doses of external ionising radiation immediately before conception were also more likely to be exposed to high doses before that time. Similarly, fathers exposed before their child was conceived were often exposed throughout their child’s life. As well as being unable to identify relevant “time windows” of exposure, the nature of the hazardous exposure, if it exists, remains unknown. Furthermore, high preconceptual doses are rare. If there is an occupational effect associated with paternal exposure to relatively high doses of radiation, the number of leukaemias that could be attributed to such an exposure in our study is small—such exposure could account for at most three of the 22 leukaemias diagnosed in almost 40000 reported children born during the 60 years 1937-96.

We thank the many people who contributed to the study, including management and workforce representatives who gave permission for the study to be conducted, industry staff who worked with us and, most importantly, the study participants themselves. We also thank Pat Ansell, Aurora Berra, Beverley Cooper, Patricia Fraser, Juliet Jain, Angela MacCarthy, Tuyet Ngyen, Margo Pelerin, Patrick Sampson, and Amanda Thomas. For supplying employment and dosimetry data, we thank Will Atkinson, Dallas Law and Keith Bromley (Atomic Energy Authority), George Sallit and Pauline Johnson (Atomic Weapons Establishments), and Keith Binks, Sheila Jones, and Les Scott (British Nuclear Fuels). Finally, we thank the scientific steering group (Eva Alberman, Tim Bishop, Martin Bobrow, and Dudley Goodhead) for continued support and guidance.

Papers p1453