Using self-assigned ethnicity, we compared the incidence of breast and gynaecological cancers between the 6 largest non-White ethnic groups in England and Whites. Overall, our findings indicate that there are considerable differences in the incidence of all 4 cancers by ethnicity; incidence rates for breast, ovarian and cervical cancer were highest among Whites, whereas the incidence of endometrial cancer was highest among Blacks. Furthermore, we found strong evidence of heterogeneity within the South Asian group, with Bangladeshis having the lowest rates of all 4 cancers.
Our finding that breast cancer incidence was lower in non-White ethnic groups compared to Whites is broadly consistent with previous studies from the UK [5, 7–9] The particularly low incidence of breast cancer among South Asians, which has been reported elsewhere [5, 7], can be largely explained by known risk factors. On average, South Asians in England have more children, are more likely to breastfeed, less likely to use HRT, much more likely to be a non-drinker, and have a lower average height than their White counterparts [19–22]. Indeed, a recent prospective cohort study of women aged over 50 found that, once incidence rates were adjusted for known risk factors, rates among South Asians were similar to those of Whites [19].
Ethnic differences were also observed within the South Asian group, with Bangladeshis having much lower rates than both Pakistanis and Indians, even after adjustment for socioeconomic status. This finding is consistent with other research [7, 23] and may be related to the higher parity, greater likelihood of breastfeeding or younger average age at first birth of Bangladeshis compared to the other South Asian groups [22–24]. Furthermore, in contrast to Indians and Pakistanis, who experienced much higher rates than their countries of origin, rates among Bangladeshis in our study were very similar to those reported in Bangladesh [2]. This suggests that Indian and Pakistani females may have adopted Western lifestyles and behaviours to a greater extent than Bangladeshi females. However, data on the prevalence of risk factors among Bangladeshis is very limited so further investigation would be needed to explain this disparity.
Moreover, contrary to expectations, we found that the rate ratio for South Asians compared to Whites was lower among under 50s compared to over 50s. Relative to older age groups, a much higher proportion of South Asians aged under 50 are UK born [25]. Therefore, we would expect the risk factors, and therefore incidence rates, for this group to be closer to those of Whites. Indeed, there have been significant falls in parity amongst South Asian women over the last 40 years (from 4 to 2.5) whereas the rate in White women has stayed fairly constant (less than 2) [22]. Although a previous study of breast cancer in ethnic groups found that rates for Bangladeshis and Whites were much closer in younger compared to older age groups, there was no clear effect of age among Indians or Pakistanis [7].
Like other UK studies, we also found lower incidence rates of breast cancer among Blacks compared to Whites [5, 7]. Again, this difference can largely be explained by known risk factors, with Blacks having more children, being younger at first birth, more likely to breastfeed, less likely to use HRT and less likely to drink alcohol [19–21]. When analysed by age, there was a marked difference in the Black-White ratio between under 50s and over 50s, a finding that has been reported in other studies from the UK [5, 9]. This is despite the fact that parity amongst blacks (about 2) has not declined over the last 40 years [22]. Studies from the US have also reported a ‘Black-White crossover’, with higher rates of breast cancer in Blacks compared to Whites in the younger age groups and the reverse pattern in older age groups [9, 26, 27]. One study, which examined ethnic differences by molecular subtype, found that this age-related difference was largely due to high rates of triple negative breast cancer among Blacks in younger age groups and high rates of HR+/HER- breast cancer among Whites in older age groups [27]. However, it is unclear what risk factors would underlie these differences.
The low rates of breast cancer among Chinese in our study have been reported elsewhere in the UK [5, 7, 28] and are consistent with international comparisons, which reveal much lower rates of breast cancer in China compared to Western countries [2, 29]. Data from the Health Survey for England reveals a high prevalence of some protective factors among Chinese, including short stature, low BMI, and relatively low alcohol consumption [21]. However, Chinese women also have had the lowest parity of all ethnic groups in England since the 1980s [22]. We might also have expected rates to be lower in older Chinese women than in younger Chinese women due to the significant fall in parity over the last 40 years (from 2.2 in 1977 to 1.3 in 2006) but our results did not show any difference by age [22].
Compared to breast cancer, very few studies have investigated the incidence of gynaecological cancers by ethnicity in the UK.As far as we are aware, this is the first study to compare the incidence of gynaecological cancers by their individual ethnic groups. ((i.e. Indian, Pakistani, Bangladeshi, Black African and Black Caribbean) as opposed to the artificially combined categories of ‘Asian’ and ‘Black’ as was done in the only previous study [5].
We observed lower rates of ovarian cancer among Blacks and South Asians compared to Whites, findings which are consistent with studies from both the UK and US [5, 30, 31]. These differences are likely to be attributed to the higher parity, longer duration of breastfeeding and lower HRT use among both these groups [19, 20, 22]. We also found evidence of intra-ethnic differences, with high incidence rates among Pakistanis and Black Africans relative to the other South Asian and Black groups. Low rates of oral contraceptive use among both these groups and low initiation of breastfeeding among Pakistanis may contribute to these higher rates [24, 32]. However, data on the prevalence of most risk factors by individual ethnic group is scarce. In contrast, rates of ovarian cancer among Chinese were similar to Whites. This is unexpected given that their rates of breast cancer (which shares several major risk factors with ovarian cancer [33]) are so low. Rates were also higher than those reported in Hong Kong, where most Chinese in the UK originate from [29]. However, the results in Chinese are consistent with them having the lowest parity of all ethnic groups in England (as discussed above in relation to breast cancer) [22].
The incidence of cervical cancer in our study was highest in Whites and results were broadly similar to those found elsewhere in the UK [5]. The particularly low rates that we observed among South Asians have previously been documented [5, 34] and may be due to the sexual behaviour of this group; although data is not available for Bangladeshis, Indians and Pakistanis tend to be older at first intercourse, have fewer sexual partners, and are less likely to be sexually active than their White counterparts [32, 35]. Similarly, incidence rates among Blacks, specifically Black Caribbeans, were lower than those of Whites. Data from both England and the US has previously revealed high cervical cancer incidence rates among Blacks relative to Whites [5, 36, 37]. However, these results are likely to have been confounded by socioeconomic differences. Indeed, before adjusting for socioeconomic status, rates among Black Africans were actually higher than those of White in our study. Nevertheless, our finding that rates were considerably lower among Black Caribbeans is somewhat surprising, especially given that there is very little difference between the number of sexual partners, average age at first intercourse and screening uptake of Black and Whites [32, 35, 38].
In contrast with the other cancers studied, Blacks, specifically Black Caribbeans, had the highest rates of endometrial cancer and we found no difference in incidence between South Asians, Chinese and Whites. Indeed, previous reports from the UK have found small or no differences in incidence or mortality between South Asians and Whites [5, 34, 39]. Nevertheless, we found strong evidence of intra-ethnic differences in the South Asian group, with rates among Bangladeshis around 50% lower than those of Indians, Pakistanis or Whites. Again, the shortage of data on the prevalence of risk factors limits our ability to explain these disparities. However, the lower prevalence of obesity, high parity, and higher initiation of breastfeeding among Bangladeshis may contribute to these differences [21, 23]. The higher incidence of endometrial cancer among Blacks has previously been reported by the NCIN [5]. Racial differences in the prevalence of obesity, which is more common in Black compared to White females, may account for some of this disparity. However, in the US, where there is also a higher prevalence of obesity among Black females [40, 41], incidence rates among Blacks are lower than those of Whites [42, 43]. Ethnic differences in the rate of hysterectomies could also contribute to these differences but, to our knowledge, there is no data available on hysterectomy rates by ethnicity in the UK.
Rates of breast, ovarian and endometrial cancer observed among the non-White ethnic groups were generally higher than their countries of origin [2]. Although this may be due to under-diagnosis or poor registration in these countries, it may also be indicative of migrants’ lifestyles and reproductive behaviour becoming more similar to that of Whites. Indeed, a study of South Asians in Leicester found that rates of breast cancer among South Asians between 1990 and 1999 increased towards those reported for Whites, presumably due to younger generations adopting more western lifestyles and reproductive behaviours [44]. Cervical cancer rates, on the other hand, were lower in our study compared to data from the countries of origin [2]. This is likely to be due to the better quality and coverage of cervical screening in this country compared to less-developed countries [45], which can allow for detection and treatment of precursor lesions [46, 47].
To our knowledge, this is the first study to compare incidence rates of breast and gynaecological cancers between the 6 biggest non-White ethnic groups in England. Previous studies have reported breast cancer incidence among these groups but were limited to a single cancer registry [7, 9]. Our use of self-assigned ethnicity was one of the major strengths of this study. This method of classifying ethnicity has a number of advantages over older systems, such as name analysis or the use of death certificates. Importantly, it allowed us to distinguish between similar ethnic groups, revealing patterns which would otherwise be concealed under the broad groupings of South Asian or Black. Furthermore, unlike the use of death certificates, it allows us to identify UK-born individuals, not just those born in other countries. It also overcomes the issue of numerator-denominator bias as the same measure of ethnicity is used for both cases (numerator) and persons at risk (denominator) [3]. Another important strength of our study is that we adjusted for socioeconomic status which is a potential confounderin studies of health and ethnicity due to the variations in deprivation between the different groups [25, 48].
One of the main limitations of this study is the lack of individual-level information available on risk factors. Population-level data on reproductive and lifestyle factors is available for the major ethnic groups [19, 20, 22, 23], allowing us to make broad ecological comparisons and generate hypotheses. However, there is very limited data for the individual ethnic groups and further investigation is needed in this area. Another limitation is the proportion of missing ethnicity data. Information on ethnicity was missing in approximately 20% of cases. However, this figure is much lower than previous studies conducted on earlier data [7, 9] and assigning ethnicity values to missing data using multiple imputation in our sensitivity analysis made no difference to our results. While the results from the imputation analyses are reassuring, they should be interpreted with caution. Multiple imputation is based on the assumption of missing at random. If this assumption does not hold, (i.e. if persons from ethnic minorities are less likely to report their ethnicity), the results may be biased [49].