This study provided a unique opportunity of investigating time trends in educational disparities in breast cancer mortality in France: the study population is a large, representative sample and individual data on socioeconomic status and specific cause of death were obtained from the same source for all subjects, irrespective of their vital status. Thus, these results are not affected by the numerator/denominator bias of studies in which educational details are derived from different sources for the deceased and the living (Kunst et al, 1998
Coding for educational level in the census changed slightly in 1990. Some degrees were classified as professional until 1982, being grouped with high school and higher education in 1990. This may lead to a slight underestimation of the educational inequalities level for the last period, but as it concerns only a small proportion of degrees it probably does not much bias the results.
This study throws new light on both social inequalities in breast cancer mortality and their time trends in France. Our analyses by age at death and birth cohort gave consistent results. We found that educational differences in breast cancer decreased between 1968 and 1996 from a positive association in the 1970s until no association remained in the 1990s. This association disappeared earlier among younger women; it was found among women born before 1925, but not among those born after 1925.
Two studies investigating time trends in social inequalities by broad causes of death found that the positive association between socio-economic status and breast cancer mortality remained stable between 1959–1972 and 1982–1996 (Steenland et al, 2002
) or appeared between 1981 and 1991 (McLoone and Boddy, 1994
). Two other studies in Finland and the US focused on breast cancer mortality (Wagener and Schatzkin, 1994
; Martikainen and Valkonen, 2000
). The Finnish study used individual data on education, whereas the American study used an ecological measure; both showed a decrease in social disparities. Our results are in agreement with these findings. During the last few decades, educational level among women strongly increased in all industrialised countries (Martikainen and Valkonen, 2000
). The above two studies did not control for this change, but our results show that even after taking this into account (by the use of RII as the measure of social inequalities), social disparities decreased over time.
There are several possible explanations for our findings, but primarily changes in the social distribution of breast cancer risk factors associated with education during the study period. Reproductive factors, diet, alcohol consumption, excess weight, physical activity and genetic factors (family history of breast cancer) are all associated with breast cancer incidence (Hulka and Stark, 1995
; Hankinson et al, 2004
). Data on these risk factors were not available; so no conclusion can be drawn on their contribution to the observed time trends. Nevertheless, a French study showed that parity decreased with the year of birth for women born between 1917 and 1949, and this was more pronounced among women with lower education levels (Daguet, 2000
); this may partly explain the decrease over time of educational disparities for breast cancer. Factors associated with cancer survival may also contribute. We observed that, within each age group, educational disparities diminished over time. It may be a consequence of better prevention and treatment in recent years, which mostly benefited women with higher education levels, as evidenced by their higher screening rates (use of mammography and breast examination) (Heck et al, 1997
; Katz et al, 2000
; Gupta et al, 2003
; Remontet et al, 2003
). This improves the relative position of the educated women in terms of mortality and thereby diminishes the social disparities in breast cancer mortality. In France, systematic screening with mammography began in 1990. Although it has not yet been evaluated in France until now, the introduction of systematic screening is probably too recent to have had an impact on educational disparities during the last period (1990–1996).
There is also some evidence that risk factors for breast cancer may differ for premenopausal and postmenopausal cancers (Hulka and Stark, 1995
). Family history of breast cancer is particularly relevant for premenopausal cancers, whereas reproductive and behavioural risk factors are generally more important for postmenopausal cancers. Breast cancer mortality before age 50 may be considered as due to premenopausal cancers. Differences in risk factors according to menopausal status could partly explain why we did not find any association between education and mortality in this age group after 1975, whereas we observed that educational disparities among older women were more pronounced. The literature on educational disparities according to menopausal status is particularly scarce, but one study found a slightly steeper gradient among postmenopausal women (Braaten et al, 2004
). We found no association between education and breast cancer mortality among women born after 1925. Breast cancer deaths occurring among this birth cohort who were aged 35–50 in 1975, 35–57 in 1982 and 35–65 in 1990 were probably not all premenopausal cancers. Thus, we cannot attribute this lack of social inequalities among women born after 1925 to the lack of an association between socioeconomic status and premenopausal breast cancer mortality. The only study of social inequalities by birth cohort found small educational differences in breast cancer mortality among women born after 1935 (Martikainen and Valkonen, 2000
The findings show that the positive association between education and breast cancer mortality progressively disappeared between 1968 and 1996, and was not observed among women born after 1925. An important question is whether the changes are indirect, and due to changes in the prevalence of breast cancer risk factors associated with education that we could not investigate.