Childhood cancer incidence by ethnic group in England, 2001–2007: a descriptive epidemiological study

In 2009, Cancer Research UK (CRUK) published Cancer Incidence and Survival By Major Ethnic Group for England 2002–2006 [1], linking incidence and mortality data from cancer registries with (self-assigned) ethnicity from the Hospital Episodes Statistics (HES) database. This methodology allowed a much more accurate estimation of outcomes by ethnic group and confirmed differences in incidence and survival in many of the different cancer types, with CRUK concluding that these differences needed ‘investigating further and the analyses extended’.

We have since published a series of papers to do that by looking more closely at individual ethnic groups and their differences in cancer incidence. South Asians and Blacks are not homogenous groups, with the subgroups within these broad categorisations having differing religious, social and cultural practices. We thus analysed cancer incidence in gastrointestinal [2], haematological [3], thyroid [4], breast and gynaecological [5], urological [6] and CNS [7] malignancies nationwide, looking individually at the difference between British Indians, Pakistanis, Bangladeshis (‘South Asians’), Black Africans, Black Carribeans (‘Blacks’) and Whites. These consistently show differences in incidence between the ethnic groups in many cancers; interestingly, they also suggest that these differences – between British Whites and ethnic minorities - can become less marked in some cancers over time [2], in keeping with previous studies in migrant populations [8] and suggesting possible environmental risk and protective factors where such patterns are observed over the space of a few generations.

Whilst some of these differences can be accounted for through known risk factors, there are many for which we do not currently have any good explanation. Thus, accurately confirming these ethnic differences (and related outcomes) through linked data, and using self-assigned ethnicity as the current most accurate measure of ethnicity [9, 10], allows not only for targeted public health spending and interventions, but is also a first step in attempting to identify potentially modifiable risk factors.

Beyond the first year of life, cancer is the commonest cause of death in childhood (ages 0–14) in England and Wales [11]. Whilst the CRUK report did not study this, here we consider for the first time the nationwide data for childhood cancer (2001–2007), using self-assigned ethnicity data and widening the analysis to include all of the above mentioned ethnic subgroups.

The methods used in this study were broadly the same as those described in our previous studies [2‐7, 12].

Demographic information for children in England from the 2001 Census is presented in Table 1. The total childhood population in England was 9,277,814 of which the majority (84.2%) were White.

There is a greater proportion of older children amongst Whites, Indians, Black Carribeans and Chinese, with the reverse being seen in Pakistanis, Bangladeshis and Black Africans. Levels of deprivation also differed with the majority of Pakistanis, Bangladeshis, and Blacks having low incomes and the remaining ethnic groups being mostly middle or high income.

The majority of children were UK born, though the proportion varies between different ethnic groups from 68% in Black Africans to 93% in Black Carribeans and Indians.

Analysing nationwide data for childhood cancer, we have presented results in this paper that suggest an overall increased risk of childhood cancers in Pakistani and Black African children relative to White children. We were also able to further assess the major childhood cancers and their incidence within self-reported ethnic groups. Here, we confirmed the well described [24‐28] increased risk of leukaemia and lymphoma in South Asian children, but for leukaemia also show this being due to the greater risk in Pakistani children in particular. In contrast to our findings in Indian children in Leicester, South Asian children appeared to have a lesser risk of CNS cancers. This has also been found in previous studies [24, 27, 29, 30], but these studies were underpowered and did not reach statistical significance, nor provide evidence for the lower risk of CNS cancers in Pakistani children compared to Indian children. Finally, an increased risk of ‘other solid tumours’ was observed in Black African children, likely driven by the previously described excess of renal tumours in this ethnic group [31, 32], though in this study we did not have sufficient numbers to estimate the relative risk.

We have previously discussed [12] how ethnicity is likely a proxy for genetic and/or environmental factors that might modify cancer risk, and how varying rates of cancers between ethnic groups may therefore be explicable through exploring the (differing) prevalence of putative risk/protective factors between ethnic groups. Where data for different ethnic groups could be found, some such factors are presented in Table 3, and discussed further below.

High birthweight has been associated with an increased risk of leukaemia (and possibly non-leukaemia cancers in older - ≥3 years old – children) [33]. Similarly, advancing maternal age has also been associated with a small increased risk (<10%) [34] in all groups of childhood cancer – leukaemia, lymphoma, CNS - analysed in this study. The above and other factors, such as maternal alcohol consumption in pregnancy [35], and maternal [36, 37] and paternal smoking [38, 39], all of which been shown to be associated with an increased childhood cancer risk (albeit inconsistently and to varying degrees for different cancers and subtypes) are all generally of greater prevalence in British Whites. Yet, our main findings are those of an increased cancer risk overall, and in leukaemias and lymphomas in particular, in some South Asian and the Black African ethnic minority groups.

As seen in Table 1, whilst a greater proportion of these groups in whom we observed a higher risk of leukaemias (Pakistanis) and lymphomas (South Asians and Black Africans) are from a lower income domain (and this study has adjusted for deprivation), recent large representative population based studies have not observed an association of deprivation with leukaemia or lymphoma subtypes [40, 41].

The relatively greater prevalence of HIV in Black Africans (in whom HIV exposure is mainly through sex between men and women [42]) is likely driving the increased risk of childhood lymphoma observed here and in other studies [43‐45].

The one group of cancers in which a reduced risk relative to British Whites was observed (RR = 0.71) was in CNS cancers in South Asian children (apparently driven by a 32% lesser risk in Pakistani children). This finding is in keeping with previous UK studies (referenced above), many of which were in communities wherein there are large Pakistani populations, and showed a similar (though non-significant) reduced risk in South Asians relative to non-South Asians.

There are few well established risk factors for childhood CNS cancers [46‐48]. Of these are a number of hereditary syndromes, which given the higher rate of consanguineous marriage in Pakistani families [49], one might expect a similarly increased risk of CNS cancers. However, such syndromes are thought likely to contribute to relatively few cases [50]. Asthma, or atopy more generally, is a more prevalent, proposed protective factor in CNS cancers [51, 52]. However, its prevalence does not appear to be markedly different across different ethnic groups [53] (Table 3).

This study has many of the strengths of our previous studies, namely the use of self-assigned ethnicity as a more accurate measure of ethnicity, as well as the same method being used for both numerator (Cancer Registry and HES) and denominator (Census), and the ability to separate large heterogeneous ethnic groupings (e.g. South Asian, Black) into more ethnically similar subgroups.

A further additional strength of this study relative to our analysis of childhood cancer in Indian and White children in Leicester was the use of national data, with this much greater sample size and number of outcomes allowing for greater power and precision in our estimates. We were able to adjust for age, sex and deprivation (all potential confounders in studying the association between ethnicity and cancers), and indeed also present results by sex. Using national data, where the method of ascertainment of cases is similar across the country also allows for a more accurate comparison between ethnic groups relative to those studies which compare rates of disease in groups in different countries. This is, to our knowledge, the first national study of childhood cancer incidence rate ratios between ethnic groups using self-assigned ethnicity. Additionally, there was little missing ethnicity data (Table 2) in these cancers, markedly lower than in our previous studies and other studies which have used HES data in combination with other methods [30].

Limitations remain however, in this being a population level study without information on individual exposures. Further, despite our presenting results by smaller, more homogenous ethnic subgroups, there remain within these groups a degree of heterogeneity, e.g. with Black Africans having a number of countries of origin, and similarly with Indians and Pakistanis originating from a number of provinces and states, with the cultural and genetic diversity that results in different ethnic groups.

As we have previously noted [6], we considered the ‘White’ classification to be ‘British White’ - though there would have been ‘Irish White’ and ‘Other White’ present in the ‘White’ classification (until 2003). However, these would have been very few (4% in the 2001 census), and unlikely therefore to have affected the results for British Whites. Finally, despite our use of self-assigned ethnicity as the current best measure of ethnicity, there remains a discordance – more so in ethnic minorities - between HES ethnicity recording and self-assigned ethnicity and there is an ongoing need to improve the accuracy of this data [54].

Improvements in data collection and linkage of databases in recent years have permitted a more detailed and accurate study of ethnicity as a possible risk or protective factor in a number of different diseases. Initial descriptive studies such as this highlight associations between ethnicity as an exposure and outcomes such as childhood cancers; whilst it is not yet of course possible to draw conclusions regarding correlation, the awareness of these differences between ethnic groups based on high quality data allows for better public health planning and targeted initiatives, and the development of further research to aim to understand why these differences might exist, potentially giving rise to individual level, translational research [55].