Introduction
Extensive mammographie density reduces the sensitivity of traditional screen-film mammography thus making the detection of breast cancer by this method more difficult1. Extensive mammographie density is also an independent risk factor for breast cancer2,3.
There is evidence showing urban-rural differences in breast cancer incidence and mortality, both generally being higher in urban areas4,5. Area of residence is related to a number of socio-economic factors which have been found to be associated with breast cancer risk. The increased risk with respect to higher socioeconomic status has been found to be attributable to reproductive and hormonal factors, body habitus and exposure to mammography screening6. Mechanisms of risk with respect to breast density, however, are not well understood. If density was found to vary with area in a similar manner to breast cancer incidence, this would give further understanding of the hormonal aetiology, and would, therefore, generate hypotheses on prevention. It would also indicate populations which might be in need of more intensive preventative or early detection interventions. There is anecdotal evidence that women living in rural areas have less dense breasts than women living in urban areas. There is, however, no empirical data published on this subject. In this article, we compare breast densities of women living in London with women resident outside the capital.
Materials and methods
The Princess Grace Hospital is an independent hospital in central London. Among other services, it provides breast care, including digital mammography screening. While many of the patients are resident in London, the hospital receives referrals from all over the UK, and from overseas.
This study was approved by the Medical Advisory Committee of Princess Grace Hospital.
Power calculations showed that 300 women from London and 600 from outside London would give approximately 90% power for a comparison of 20% fatty replaced in the former group with 30% in the latter. Amongst 7297 women having digital mammographic examinations at the Princess Grace Hospital in London in 2006, we randomly selected, by computer, 318 examinations in screen negative women resident in London, as defined by London postcodes, and 656 from women outside the capital, a total of 972. Of the 163 London postal sub-districts, 91 (56%) were covered in our sample. Reflecting the geographic location of the hospital, the majority (61%) were from west and south-west London postal districts. We did not impose a time limit on residence.
All mammograms were high quality two-view images, soft copy reported on high resolution monitors. Image quality was not a factor in case selection before randomization, and because all images were adequate, none were rejected during the density estimation. Both views were used for estimation of density, as it is known that the craniocaudal view tends to give higher density estimates than the mediolateral oblique7. The density was estimated by a single, experienced radiologist (one of the authors: NMP), in ignorance of the area of residence, using the four standard density grades according to the American College of Radiology Breast Imaging Reporting and Data System (BIRADS). This system classifies density into four patterns: 1: < 25% dense (almost entirely fatty); 2: 25-50% dense (scattered fibroglandular density); 3: 51-75% dense (heterogeneously dense) and 4: > 75% dense (extremely dense). Area of residence was classified as London or elsewhere based on the postcode of residence. Where density differed between the two breasts, the higher estimate was chosen.
In addition to age, density and area of residence, we did not have access to data on breast cancer risk factors.
The association between the ordered dense categories and area of residence was assessed using ordered polychotomous logistic regression, fitting a common odds ratio estimate to each of the three gradations of the four density categories, analogous to a trend test8. secondary analyses were stratified by age and by fitting a conventional logistic regression in each age group for BIRADS categories 2-4 combined, compared to category 1.
Results
Table 1 shows the study population by area of residence and BIRADS density category. There were 318 (33%) women resident in London and 654 (67%) resident outside of London. Fewer of the London residents were classified as having almost entirely fatty breasts (21% compared to 26%) and more had extremely dense breasts (24% vs. 19%). The difference in density frequencies was significant (p = 0.04, logistic trend test). After adjusting for age in the polychotomous logistic regression, the densities of the London group remained significantly higher than the non-London (OR = 1.32, 95% CI 1.04-1.70, p = 0.02).
The mean age of the women was 55 years in each residential group (SDs 10 for the London group, 11 for the non-London), and overall ranged from 29 to 87 years. Age distributions of the two residential groups are shown in Table 2. There was no significant difference in age by area of residence.
The most striking difference by area of residence with respect to density was the lower percentage in London of BIRADS category 1, less than 25% density (Table 1). We, therefore, fitted conventional logistic regression models within each age group for density of 25% or more compared to less than 25%. Results are shown in Table 3. Odds ratios were elevated in all age-groups for women resident in London compared to women resident outside the capital and this was significant for the age group 45-54 years (OR = 2.22, 95% CI 1.05-4.68, p = 0.035) when adjusted for the residual effect of age within age groups. This heterogeneity by age was not significant, however.
Discussion
We found that women resident in London had significantly higher breast density than women resident outside the capital, after adjusting for age. The major limitation of the study is that we do not have confounding factor data at individual level. Breast density has been shown to be related to breast cancer risk factors using both quantitative and qualitative measures of density9-12. It is inversely related to the risk factors of age and body mass index11. It has also been shown to be related to ethnicity9. Since we do not have risk factor data other than age, the major question raised by this study is: can the increased density in London women be attributed entirely to known risk factors other than area of residence, and if so which factors? It seems unlikely that it is due to lower parity among London residents, since fertility rates are similar in London to those in England as a whole13. It is noteworthy that breast density is inversely related to body mass index, and that the Health Survey for England reported that women resident in London had lower body mass index than women living outside of the capital14. This is consistent with our results but is an ecological observation. Study of place of residence, breast density and other risks factors at an individual level is indicated.
In age-specific analyses, the increased density in London women was significant only for the age group 45-54 years. Since these mammograms were taken in 2006 after the use of hormone replacement therapy had been in decline for some years, this is unlikely to be due entirely to differences in hormone replacement therapy use. Figure 1 shows the age-specific incidence of breast cancer in England and Wales as a whole and in London, in the years 1999-2001 inclusive15. At younger ages, incidence is lower in London, but at about age 45 years, the rates become approximately equal and remain so thereafter. It is not clear why incidence is lower in London at younger ages. Also, the catch-up in middle age may or may not be attributable to the increased levels of density in London in women at ages 45-54 years, but it is at least suggestive. It is possible that the lower incidence in London at earlier ages is due to confounding with ethnic factors, or possibly to under-registration at young ages.
Our work adds to this in showing that it is also related to area of residence. It would be interesting to see if density is higher in urban areas compared to rural or suburban areas in other countries.
Density is modifiable by dietary habits and drug therapy11,12. The important question which remains is whether lowering a subject's breast density by whatever means confers a reduction in that subject's risk of future breast cancer, or whether density is a marker for damage which has already been done. Our results, at any rate, suggest a higher level of density in urban areas. If this is found to be the case more generally, it indicates both higher risk and a greater challenge for mammographic screening in urban areas. It is noteworthy that screening uptake is 62% in London compared with the national average of 75%16.
Further research plans in this project include having density measured as a continuous percentage using the Cumulus interactive threshold software, and further classification of area of residence outside London as rural, suburban and urban. In addition, of course, we plan to collect individual data on the potential confounding factors, including body mass index, parity and use of exogenous hormones. It is also planned to consider duration of residence.
Conclusion
Pending further work, our preliminary results suggest that women in London have greater breast density than women resident elsewhere, and that this is most pronounced in the peri-menopausal age group 45-54 years. It might be prudent in the meantime, bearing in mind the limitations of the present study, to ascertain the cause of the phenomenon and to optimize uptake and quality of breast cancer screening in the capital.
Acknowledgements
Declarations of interest: This study was funding by the Princess Grace Hospital, London (part of HCA International headquartered at Nashville, TN, USA).
We would like to thank Dr Ashu Sehgal at the Thames Cancer Registry for supplying us with breast cancer incidence rates for London, and Miss Saima Ayyaz at the Wolfson Institute of Preventive Medicine for her administrative assistance.
Aside from institutional support provided by HCA Healthcare, no other company supported the study or the publication of this article. Some of the authors have given talks, attended conferences and participated in trials and advisory boards sponsored by various pharmaceutical companies.
(C) 2008 Current Medical Research and Opinion. via ProQuest Information and Learning Company; All Rights Reserved
