Background Inequalities in infant mortality in England and Wales remain of concern. Patterns of risk for preterm babies (over 50% of infant deaths) are poorly understood, and may differ by cause of death and degree of prematurity. We aimed to describe ethnic and social variation in cause-specific infant mortality of preterm babies by level of prematurity, and examine the extent to which disadvantage might mediate ethnic variation, or ethnicity explain social variation. Methods The Office for National Statistics linked birth and death registrations with other routine data for singleton live births at gestational age 24–36 weeks in England and Wales 2006–2012. Within three gestation categories (24–27, 28–31, 32–36 weeks), we fitted Poisson regression models with robust standard errors (adjusted for potential confounding by gestation week, birth year, and gender), relating ethnicity (9 groups) and/or area deprivation (IMD quintiles) to risk of infant death from congenital anomalies, immaturity-related conditions, or all other causes. Results There were 2 56 142 births and 6480 deaths (26% from congenital anomalies, 52% immaturity-related). Among 24–27 week births (5% of preterm babies, but 47% of those who died in infancy), infants of all minority ethnic groups had lower risk of immaturity-related death than White British, the lowest rate ratios being 0.65 (95% CI 0.51 to 0.83) for Black Caribbean, 0.76 (0.66–0.87) for Black African, and 0.76 (0.61–0.94) for Indian. Among 32–36 week births, infants of all minority groups had higher risk of death from congenital anomalies than White British, the highest rate ratios being 4.86 (4.10–5.77) for Pakistani, 3.12 (2.28–4.28) for Bangladeshi, and 2.21 (1.71–2.87) for Black African. Risks of death from congenital anomalies and 'other' causes increased with disadvantage, rate ratios comparing the most with the least deprived quintile being respectively 2.01 (1.60–2.51) (attenuated to 1.54 (1.22–1.94) by adjustment for ethnicity) and 2.07 (1.57–2.73). Except for social variation in death from congenital anomalies, adjusting disadvantage models for ethnicity, or vice versa, made little difference. Conclusion Ethnic variation in infant mortality following preterm birth is driven by contrasting patterns of death from immaturity-related conditions in 24–27 week babies, and congenital anomalies in 32–36 week babies. Social variation is driven by deaths from congenital anomalies (perhaps in part explained by ethnicity) and 'other' causes. We found no evidence that ethnic variation was mediated by disadvantage. Further work is needed to establish reasons for ethnic variation in death from congenital anomalies. Future research should examine biological causes of very preterm birth.

Background About 8% of deaths under age 1 year in England and Wales are unexplained, the cause of death being recorded as sudden infant death syndrome (SIDS, 60%) or unascertained (40%). Typically, unexplained infant death (UID) occurs during sleep, perhaps triggered by unsafe sleep environments (prone/side position, soft bedding etc.). Other risk factors include preterm birth, which may increase vulnerability. Incidence of both preterm birth and (in an analysis of data from 2005) SIDS has been relatively high for Black Caribbean babies. We aimed to evaluate recent ethnic variation in UID, test the hypothesis that it might be mediated by preterm birth, and assess several other potential explanatory factors. Methods For singleton live births at gestation 22+weeks in England and Wales during 2006–2012, the Office of National Statistics provided a file linking birth and death registrations with other routine data. Variables included ethnicity (11 groups) and six covariates: gestational age, area deprivation (IMD quintile), gender, mother’s age at delivery, mother’s country of birth (UK-born or not), parental status at birth registration (within marriage, jointly registered by both parents at same address, joint registration with different addresses, sole registration). Using Stata, we fitted a sequence of logistic regression models for UID, starting with ethnicity and adding factors in the order in which they reduced the likelihood ratio test statistic for ethnic variation. Results There were 4.6 million births and 1559 UID events (0.34 per 1000). Risk varied by ethnicity (p<0.001), being low for Black African, White non-British, and all South Asian groups, but high for Black Caribbean and two Mixed Black groups (African-White and Caribbean-White). Unadjusted odds ratios relative to White British ranged from 0.38 (95% CI 0.24–0.60) for Indian to 1.73 (1.21–2.47) for Black Caribbean. All covariates were associated with UID risk. When jointly adjusted for two factors (parental registration status and mother’s country of birth), the ethnic variation was not statistically significant (p=0.6). Other factors, including gestational age, made very little difference. Results were robust to sensitivity analysis for potential sparse-data bias. Conclusion This large population-based study found substantial ethnic disparity in risk of UID, which was not attributable to preterm birth, maternal age or area deprivation. Parental registration status and mother’s country of birth jointly accounted for the variation, suggesting mediation by cultural factors. Better understanding of infant care practices in low-risk ethnic groups might lead to more effective prevention of UID in the general population.

On 1 November 1983 a programme “Windscale – The Nuclear Laundr y” made by the Yorkshire Television A media report in 1983 drew attention to high levels of childhood leukaemia around the Sellafield nuclear reprocessing plant. This prompted investigations around other nuclear installations, some of which suggested other “clusters”, though Sellafield remained the most striking. Many studies over more than 30 years have investigated possible reasons for such clusters. Inevitably attention was first directed at radiation linked with activities at the plant. However, it was found that doses from accidental and planned releases were too low to account for the observed levels of childhood leukaemia. Various other mechanisms involving radiation have been investigated and have also been discounted. While no clear explanation for the Sellafield cluster has been found, perhaps the most plausible remaining hypothesis involves “population mixing” in which an infection is spread to susceptible individuals and, in rare cases, results in leukaemia.

Greater adiposity and height have been associated with increased risk of haematological malignancies. Associations for disease subtypes are uncertain. A cohort of 1.3 million middle-aged UK women was recruited in 1996–2001 and followed for 10 years on average. Potential risk factors were assessed by questionnaire. Death, emigration, and incident cancer were ascertained by linkage to national registers. Adjusted relative risks were estimated by Cox regression. During follow-up, 9162 participants were diagnosed with lymphatic or haematopoietic cancers. Each 10 kg m−2 increase in body mass index was associated with relative risk of 1.20 (95% confidence interval: 1.13–1.28) for lymphoid and 1.37 (1.22–1.53) for myeloid malignancy (P=0.06 for heterogeneity); similarly, Hodgkin lymphoma 1.64 (1.21–2.21), diffuse large B-cell lymphoma 1.36 (1.17–1.58), plasma cell neoplasms 1.21 (1.06–1.39), acute myeloid leukaemia 1.47 (1.19–1.81), and myeloproliferative/myelodysplastic syndromes 1.32 (1.15–1.52). Each 10 cm increase in height was associated with relative risk of 1.21 (1.16–1.27) for lymphoid and 1.11 (1.02–1.21) for myeloid malignancy (P=0.07 for heterogeneity); similarly, mature T-cell malignancies 1.36 (1.03–1.79), diffuse large B-cell lymphoma 1.28 (1.14–1.43), follicular lymphoma 1.28 (1.13–1.44), plasma cell neoplasms 1.12 (1.01–1.24), chronic lymphocytic leukaemia/small lymphocytic lymphoma 1.23 (1.08–1.40), and acute myeloid leukaemia 1.22 (1.04–1.42). There was no significant heterogeneity between subtypes. In middle-aged women, greater body mass index and height were associated with modestly increased risks of many subtypes of haematological malignancy.

BACKGROUND:Inclusion in clinical trials is generally viewed as best practice for most newly diagnosed childhood cancers, but the impact on population-based survival has rarely been examined.PATIENTS AND METHODS:The population-based data were analysed for 25 853 children (66% of all registered childhood cancers) diagnosed in Britain during 1978-2005 with acute lymphoblastic leukaemia (ALL), acute myeloid leukaemia (AML), Hodgkin lymphoma, non-Hodgkin lymphoma, medulloblastoma, neuroblastoma, Wilms tumour, hepatoblastoma, osteosarcoma, Ewing sarcoma, rhabdomyosarcoma and germ-cell tumours. The Kaplan-Meier survival curves were compared by log-rank tests. Time trends were analysed by Cox regression. Separate analyses were done for children with ALL, medulloblastoma and neuroblastoma according to clinically relevant age thresholds.RESULTS:Survival increased significantly during 1978-2005 for every diagnostic category; the annual reduction in risk of death ranged from 2.7% (rhabdomyosarcoma) to 12.0% (gonadal germ-cell tumours). Survival increased steadily between trial eras for ALL (age 1-14 years) and neuroblastoma (age 1-14 years), but changed little since the mid-1980s for medulloblastoma (age 0-2 years), osteosarcoma or Ewing sarcoma.CONCLUSIONS:Changes in survival between trial eras parallel those reported by the relevant clinical trials. The increasing level of participation in trials, facilitated by the organisation of specialist care, has underpinned the substantial improvements in survival seen at the population level.

Sir, We thank you for giving us the opportunity to respond to Dr Franceschi’s (2012) letter. We reported consistently higher childhood leukaemia incidence rates in more affluent communities within England and Wales in each of the three decades up to 2005, and discussed several possible explanations (Kroll et al, 2011). Dr Franceschi queries our interest in the possibility of uneven diagnosis, and suggests that it would be of interest if the effects of adjustment by maternal parity and/or maternal age could be reported. It is true that the relationship of childhood leukaemia to the socioeconomic measure used in our study (quintiles of the Carstairs deprivation index) might have been attenuated if the analysis had been adjusted by maternal parity and/or maternal age (or any other factor related to socioeconomic status). We were unable to make such adjustments because this was a study of incidence in the whole childhood population, not a case–control study, and birth records were not available for all registered cases. However, we note that, strictly, these factors are not ‘known to influence childhood leukaemia risk’, as Dr Franceschi implies, but are known to be associated with it; the explanation is unknown. The study mentioned by Dr Franceschi (Dockerty et al, 2001) was included in the systematic review that we cited (Poole et al, 2006), and was therefore not discussed individually in our paper. This study differed from ours in several respects. It was a case–control study for the diagnosis period 1968–1986, restricted to children for whom birth records were available, and using a deprivation score derived from address at birth, rather than address at diagnosis. Nevertheless, with increasing deprivation there was a statistically significant decreasing trend in risk of acute lymphoblastic leukaemia, the major subtype in children (Table 5, Dockerty et al, 2001). We certainly did not mean to suggest that British paediatricians discriminate in any way in the care they provide. Rather, we suggest that recognition of leukaemia as a potential underlying cause of non-specific symptoms might be uneven: for example, in poorer communities, provision of primary care may be less generous, and parents may be younger and less well-educated. Thus, for example, under-diagnosis might contribute to the associations with maternal parity and maternal age that Dr Franceschi mentions. A further study (Kroll et al, 2012) uses clinical data to examine this possibility in detail.

Background: Previous research suggests associations of lower alcohol intake and higher tobacco consumption with increased risks of haematological malignancy. The prospective Million Women Study provides sufficient power for reliable estimates of subtype-specific associations in women. Methods: Approximately 1.3 million middle-aged women were recruited in the United Kingdom during 1996–2001 and followed for death, emigration and cancer registration until 2009 (mean 10.3 years per woman); potential risk factors were assessed by questionnaire. Adjusted relative risks were estimated by Cox regression. Results: During follow-up, 9162 incident cases of haematological malignancy were recorded, including 7047 lymphoid and 2072 myeloid cancers. Among predominantly moderate alcohol drinkers, higher intake was associated with lower risk of lymphoid malignancies, in particular diffuse large B-cell lymphoma (relative risk 0.85 per 10 g alcohol per day (95% confidence interval 0.75–0.96)), follicular lymphoma (0.86 (0.76–0.98)) and plasma cell neoplasms (0.86 (0.77–0.96)). Among never- and current smokers, higher cigarette consumption was associated with increased risk of Hodgkin lymphoma (1.45 per 10 cigarettes per day (1.22–1.72)), mature T-cell malignancies (1.38 (1.10–1.73)) and myeloproliferative/myelodysplastic disease (1.42 (1.31–1.55)). Conclusion: These findings confirm and extend existing evidence for associations of subtypes of haematological malignancy with two common exposures in women.

Increases in recorded childhood cancer incidence are widely reported, but do not necessarily represent real increases in risk. Time trends might conceal underlying steps caused by changes in diagnosis and registration procedures. Using records from the National Registry of Childhood Tumours 1966–2005 (N=54 650), the age-sex-standardised rate for residents of Great Britain aged under 15 years was calculated by individual year of diagnosis for each cancer subtype, and the average annual percentage change (trend) was assessed. The timing of assumed step changes in rate was estimated by iterative Poisson regression, and compared graphically with the approximate timing of innovations previously identified from published sources. Estimated timing of underlying steps approximately coincided with the following relevant innovations: biochemical assays, mid-1980s (hepatic and germ-cell cancer); diagnostic imaging, mid-1980s to early 1990s (intracranial/intraspinal tumours, neuroblastoma, soft-tissue sarcoma); revised cancer registration scheme, 1971 (leukaemia, bone and soft-tissue sarcoma); mandatory registration, 1993 (intracranial/intraspinal tumours, retinoblastoma, melanoma/carcinoma); cancer registration improvements, 2001 (leukaemia, renal and hepatic cancer). While the possibility of some real change in risk cannot be excluded, for many cancer subtypes the estimated timing of underlying step changes in rate appeared to correspond with changes in diagnosis or registration procedures. Childhood cancer may have been considerably under-recorded in the past.

Background: Recorded incidence of childhood acute lymphoblastic leukaemia tends to be lower in poorer communities. A ‘pre-emptive infection hypothesis’ proposes that some children with leukaemia die from infection without diagnosis of leukaemia. Various different blood abnormalities can occur in untreated leukaemia. Methods: Logistic regression was used to compare pre-treatment blood counts among children aged 1–13 years at recruitment to national clinical trials for acute lymphoblastic leukaemia during 1980–2002 ( N =5601), grouped by address at diagnosis within Great Britain into quintiles of the 1991 Carstairs deprivation index. Children combining severe neutropenia (risk of serious infection) with relatively normal haemoglobin and platelet counts (lack of pallor and bleeding) were postulated to be at risk of dying from infection without leukaemia being suspected. A deficit of these children among diagnosed patients from poorer communities was predicted. Results: As predicted, there was a deficit of children at risk of non-diagnosis (two-sided P trend =0.004; N =2009), and an excess of children with pallor ( P trend =0.045; N =5535) and bleeding ( P trend =0.036; N =5541), among cases from poorer communities. Conclusion: Under-diagnosis in poorer communities may have contributed to socioeconomic variation in recorded childhood acute lymphoblastic leukaemia incidence within Great Britain, and elsewhere. Implications for clinical practice and epidemiological studies should be considered.

Sir, We thank you for giving us the opportunity to respond to Dr Franceschi's (2012) letter. We reported consistently higher childhood leukaemia incidence rates in more affluent communities within England and Wales in each of the three decades up to 2005, and discussed several possible explanations (Kroll et al, 2011). Dr Franceschi queries our interest in the possibility of uneven diagnosis, and suggests that it would be of interest if the effects of adjustment by maternal parity and/or maternal age could be reported. It is true that the relationship of childhood leukaemia to the socioeconomic measure used in our study (quintiles of the Carstairs deprivation index) might have been attenuated if the analysis had been adjusted by maternal parity and/or maternal age (or any other factor related to socioeconomic status). We were unable to make such adjustments because this was a study of incidence in the whole childhood population, not a case–control study, and birth records were not available for all registered cases. However, we note that, strictly, these factors are not ‘known to influence childhood leukaemia risk', as Dr Franceschi implies, but are known to be associated with it; the explanation is unknown. The study mentioned by Dr Franceschi (Dockerty et al, 2001) was included in the systematic review that we cited (Poole et al, 2006), and was therefore not discussed individually in our paper. This study differed from ours in several respects. It was a case–control study for the diagnosis period 1968–1986, restricted to children for whom birth records were available, and using a deprivation score derived from address at birth, rather than address at diagnosis. Nevertheless, with increasing deprivation there was a statistically significant decreasing trend in risk of acute lymphoblastic leukaemia, the major subtype in children (Table 5, Dockerty et al, 2001). We certainly did not mean to suggest that British paediatricians discriminate in any way in the care they provide. Rather, we suggest that recognition of leukaemia as a potential underlying cause of non-specific symptoms might be uneven: for example, in poorer communities, provision of primary care may be less generous, and parents may be younger and less well-educated. Thus, for example, under-diagnosis might contribute to the associations with maternal parity and maternal age that Dr Franceschi mentions. A further study (Kroll et al, 2012) uses clinical data to examine this possibility in detail.

Record-based studies have generally reported association of higher childhood leukaemia incidence with higher socioeconomic status (SES), but recent findings are less consistent. We examined records from the National Registry of Childhood Tumours for evidence of this association in England and Wales during 1976–2005. All eligible leukaemia registrations (N=11940) were grouped by year of diagnosis in decades centred on census years 1981, 1991 and 2001 (N=3748, 3922, 4270, respectively). Using data from the census appropriate to the decade, SES for each case was measured by the child-population-weighted quintile of the Carstairs deprivation index of the census ward containing the address at diagnosis. In each decade, the age-standardised leukaemia rate in the poorest quintile was ∼90% of the rate in the most affluent. Using Poisson regression, the age-adjusted rate ratio per quintile decrease in SES was 0.96 (95% confidence interval 0.94–0.98; P<0.001 for trend) in 1976–1985, 0.97 (0.95–0.99; P=0.008) in 1986–1995 and 0.97 (0.95–0.99; P=0.009) in 1996–2005. Similar association was evident for lymphoid leukaemia, the major subgroup (N=9588 in total), but not for acute myeloid (N=1868) or other/unspecified leukaemia (N=484). Reported childhood leukaemia incidence in England and Wales continues to be higher in relatively affluent communities. Possible explanations include under-diagnosis of leukaemia in children from poorer communities, and/or association of higher SES with hypothesised risk factors, such as population mixing and delayed exposure to infection.

Completeness of ascertainment is a very important aspect of cancer registration. There is no recent published estimate for childhood cancer in Britain. We estimated completeness of ascertainment by the National Registry of Childhood Tumours for cancer diagnosed under age 15 years in residents of Britain during 2003–04. Stratified two-source capture-recapture was applied to notifications from general cancer registries (CRs) and specialist clinicians. Variation in notification patterns was assessed by logistic regression. Results were verified by cross-checking with Hospital Episode Statistics for leukaemia patients from England born in 1998 and diagnosed before 2005. CRs notified 92–96% of registrations, and specialist clinicians 93%. Notification patterns varied slightly according to registry region, age at diagnosis, diagnostic group, socioeconomic status, and whether the patient had died. Irrespective of stratification by these factors, the overall completeness estimate was 99–100% (assuming independence of sources). Estimated completeness was at least 99% within all subgroups, except for one region (Thames 98–99%) and two small diagnostic groups (germ-cell and gonadal cancer 98–99%, melanoma and non-skin cancer 97–98%). The independence assumption cannot be fully justified, as both sources used records from treatment centres. With this caveat, ascertainment of recently diagnosed childhood cancer in Britain appears to be virtually complete.

Aim: This paper describes the epidemiology and family history status of 1601 children with retinoblastoma in Great Britain diagnosed 1963–2002 and summarises the practical consequences for diagnosis and counselling of developments in molecular genetics. Methods: Incidence rates were analysed according to year of diagnosis and tumour laterality. Cases were classified as heritable or non-heritable on the basis of laterality and family history of the disease. Results: There were 998 unilateral cases, 581 bilateral and 22 of unknown laterality. Bilateral cases tended to be diagnosed at a younger age than unilateral. All bilateral cases are regarded as heritable, and 35% had a family history of the disease. 7% of the unilateral cases had a family history and are therefore heritable. Thus, at least (41%) of our cases are heritable. This is an underestimate, since these data on family history are incomplete. For unilateral cases aged below 1 year, the reported incidence rate increased significantly (p<0.0001) by about 2.5% per year; for the age group 1–4 years, the average increase was about 0.5% per year (not significant).

Background: Retinoblastoma occurs in both a heritable and a non-heritable form. In the heritable form, there is a predisposition to the development of non-ocular tumours. Objectives: To identify the types of non-ocular tumour occurring in retinoblastoma survivors and to produce estimates of risk for these tumours. Methods: We carried out a cohort study that included 1927 cases of retinoblastoma diagnosed in Great Britain between 1951 and 2004. Cases were ascertained through the National Registry of Childhood Tumours and followed up for the occurrence of non-ocular tumours using the routine notification system based on the National Health Service Central Registers in Britain. Results: Of the 1927 cases, 809 were known to have the heritable form of the disease and 1118 assumed to have the non-heritable form. 102 of the heritable and 13 of those classified as non-heritable developed a non-ocular tumour. The cumulative risk of developing such a tumour 50 years after retinoblastoma diagnosis was 48.3% (95% confidence interval: 38.1 to 59.7%) in the heritable and 4.9% (1.9 to 12.4%) in the non-heritable cases. The main categories of non-ocular tumours observed in the heritable cases were soft-tissue sarcomas (36 of which 21 were leiomyosarcoma), osteosarcoma (32), carcinoma (13), brain and central nervous system tumours (10), melanoma (9), leukaemia (4) and others (4). There were a total of 108 non-ocular tumours in 102 cases. Conclusions: There is a high risk of non-ocular tumours occurring in survivors of heritable retinoblastoma. These results have important implications for the clinical follow-up and counselling of survivors.

In 2008, the German Childhood Cancer Registry published the results of the Kinderkrebs in der Umgebung von Kernkraftwerken (KiKK) study of childhood cancer and leukaemia around German nuclear power stations. The positive findings appeared to conflict with the results of a recent British analysis carried out by the Committee on Medical Aspects of Radiation in the Environment (COMARE), published in 2005. The present paper first describes the COMARE study, which was based on data from the National Registry of Children's Tumours (NRCT); in particular, the methodology used in this study is described. Although the results of the COMARE study were negative for childhood leukaemia, this apparent discrepancy could be accounted for by a number of differences in approach, especially those relating to the distances from the power stations and the ages of the children studied. The present study was designed to match the KiKK study as far as possible. The incidence observed (18 cases within 5 km against 14.58 expected, p = 0.21) was not significantly raised. The risk estimate for proximity in the regression fitted was actually negative, though the confidence intervals involved are so wide that the difference from that reported in the KiKK study is only marginally statistically significant (p = 0.063).

In this population-based study of acute lymphoblastic leukaemia (ALL) diagnosed among children aged under 15 years in England and Wales during 1986–1995, we analysed incidence at census ward level in relation to a range of variables from the 1991 census, which could be relevant to theories of infectious aetiology. ‘Population-mixing’ measures, used as surrogates for quantity and diversity of infections entering the community, were calculated from census data on the origins and destinations of migrants in the year before the census. Incidence at ages 1–4 years tended independently to be higher in rural wards, to increase with the diversity of origin wards from which in-migrants had moved during the year before the census, and to be lower in the most deprived areas as categorised by the Carstairs index. This last association was much weaker when urban/rural status and in-migrants' diversity were allowed for. There was no evidence of association with population mixing or deprivation for ALL diagnosed at ages 0 or 5–14 years. The apparent specificity to the young childhood age group suggests that these associations are particularly marked for precursor B-cell ALL, with the disease more likely to occur when delayed exposure to infection leads to increased immunological stress, as predicted by Greaves. The association with diversity of incomers, especially in rural areas, is also consistent with the higher incidence of leukaemia predicted by Kinlen, where population mixing results in below average herd immunity to an infectious agent.

Aim: This paper describes the treatment and survival of 1576 children with retinoblastoma in Great Britain diagnosed 1963-2002.Methods: Survival rates were analysed according to period of diagnosis and tumour laterality.Results: Survival was calculated by calendar period of diagnosis, 1963-1982 and 1983-2002. For both unilateral and bilateral retinoblastoma, survival improved between the two periods. The survival curves for the two periods were significantly different: for unilateral retinoblastoma p<0.00001, for bilateral p<0.01. For unilateral cases, the estimated 5-year survival rates rose from 85% for those diagnosed in 1963-1967 to 97% for those diagnosed in 1998-2002. The equivalent rates for bilateral cases were 88% and 100%.Conclusion: Survival rates were already high at the start of the study period. They increased with changes in treatment regimens.