Breast cancer mortality trends in two areas of the province of Florence, Italy, where screening programmes started in the 1970s and 1990s

We compared breast cancer mortality rates in the period 1985–2000 in two areas of the province of Florence, Italy, where breast cancer screening programmes started in the 1970s (early screening (ES) area) and in 1990s (late screening (LS) area). The overall age-standardised mortality decreased in the whole period by 40.9% in the ES area (P<0.001), and by 11.3% in the LS area (P=0.030). Significant decreases in the ES area were detected in groups aged 45–54 years (61.1%; P= 0.018) and 65–74 years (44.7%; P= 0.049), whereas in the LS area no significant decrease was detected in any age group. The relatively low compliance in the first years of the programme in both areas, and the long enrolment period in the LS area could have reduced the effect on mortality. Our findings suggest that the drop in mortality in the ES area (41%) could be explained by both service screening and better care. The slight decrease in mortality in the LS area (11%) could be mainly due to better care. A reduction of about 30% is attributable to screening in the ES area over the period 1985–2000.

Mammographic screening has been shown to reduce breast cancer mortality rates (Vainio and Bianchini, 2002). Several studies in northern European countries (Hermon and Beral, 1996;Hakama et al, 1997;Van den Akker-van Marle et al, 1999;Blanks et al, 2000;Jonsson et al, 2001;Botha et al, 2003;Otto et al, 2003;Tabar et al, 2003) have reported a decrease in breast cancer mortality where nationwide screening programmes have been implemented. There has, however, been considerable debate concerning the relative contributions of screening and of improved therapy to the observed trends (Carnon et al, 1996;Blanks et al, 2000;Peto et al, 2000).
At the beginning of the 1970s in 23 rural municipalities of the province of Florence, Italy (early screening area (ES area) among about 70 000 resident women aged 25 or over, 18% of women of the province, based on 1991 Census), a mammographic screening programme was started; until 1989, women in the 40 -69 year age group were invited, whereas from that year onwards the target population was restricted to age group 50 -69 years. The number of mammograms performed in the ES area was 8000 -9000 per year. The efficacy of the programme on breast cancer mortality in the ES area has been evaluated for the period 1977 -1984 by means of case -control studies (Palli et al, 1986(Palli et al, , 1989, and by using early indicators of efficacy (Paci et al, 1990). After 1990, when the screening for women aged 50 -69 years started in the rest of the province, including the city of Florence (late screening area: LS area), the number of mammograms in the whole province increased to 28 000 -29 000 per year (Barchielli et al, 1999).
In this study, we have compared breast cancer mortality rates in the period 1985 -2000 in the ES and LS areas.

Sources of data
Data on breast cancer mortality for the period 1985 -2000 were obtained from the Tuscan Regional Mortality Register (Chellini et al, 2002), which collects death certificates of the residents of Tuscany. Incident breast cancer cases for the period 1985 -1999 were obtained from the Tuscan Cancer Register, a member of the International Association of Cancer Registries (Parkin et al, 2002).
In order to calculate the estimated percent change (EPC) in rates for the whole period of observation, trends in mortality and incidence based on annual data were examined using a log-linear regression model, and the year of death as a continuous variable.
In addition to visual inspection, we measured EPCs in rates over the whole period (Table 1). In the period 1985 -2000, the overall age-standardised mortality decreased by 40.9% in the ES area (Po0.001), and by 11.3% in the LS area (P ¼ 0.030). Significant decreases in the ES area were detected in age groups 45 -54 years (61.1%; P ¼ 0.018) and 65 -74 years (44.7%; P ¼ 0.049). In the 75 -84 age group, the decrement was 45.2% (P ¼ 0.053). In the LS area, no significant decreases were detected in any age group.
In the period 1985 -1999, the overall age-standardised incidence increased by 12.9% in the ES area (P ¼ 0.190), and by 14.2% in the LS area (Po0.001) ( Table 1). In the ES area, no significant increases or decreases were detected in any age groups; in the LS area, where the prevalence screening was carried out in the period 1990 -1994, significant increases were detected in groups aged 55 -64 years (21.0%; P ¼ 0.009) and 85 years and over (45.1%; P ¼ 0.012).

DISCUSSION
Decreases in breast cancer mortality have been variously reported following the introduction of mammographic screening on a regional level (Tornberg et al, 1994;Quinn and Allen, 1995;Garne et al, 1997;Barchielli and Paci, 2001;Broeders et al, 2001;Tabar et al, 2001;Duffy et al, 2002). Randomised prospective trials indicate that benefits in terms of cumulative breast cancer mortality start to emerge 4 -10 years after randomisation (Nyström et al, 2002). We observed a similar trend in the mortality rates in the two areas up to 1993, although the mortality Late screening area Age-adjusted rates (×100 000) Early screening area  rates in the ES area were lower than those in the LS area. Afterwards a marked fall occurred in ES area, whereas only a slight decrease was observed in the LS area. The relatively low compliance in the first years of the programme in both areas (about 55%) and the long enrolment period (on average 2.5 years) in the LS area (Paci et al, 2002a, b) could have diluted the effect so that a decrease in mortality rates was observed only after a longer interval.
Our findings suggest that the drop in mortality in the ES area (41%) could be explained by both service screening and better care. The slight decrease in mortality in the LS area (11%) could be explained by better care, while service screening of the period 1990 -2000 in the LS area could have been responsible for only a small number of deaths occurring in the target population. In fact, it has been estimated that in the city of Florence, the breast cancer mortality reduction at ages 50 -74 years attributable to screening was 3.2% over the period 1990 -1999 (Paci et al, 2002a). Thus, allowing a 11% reduction due to better care, a reduction of about 30% (41 -11%) is attributable to screening in the ES area over the period 1985 -2000. These findings are consistent with the mortality reduction of about 25% in screened populations shown by randomised trials (Vainio and Bianchini, 2002).
The decrease in mortality at ages 45 -54 years in the ES area could be related to the involvement in mammographic screening of women aged 40 -49 years in the ES area until 1993. In 1985-1993 of mammograms in the ES area (about 2700 per year) were performed on women aged 40 -49 years. No mortality reduction would be expected from screening in the younger age group (35 -44 years) or in the 85 years and over age group.
In evaluating these results, it must be borne in mind that the ES area, where the incidence rates were lower, was a less urbanised area during the period of observation. The significant increase in incidence recorded only in the LS area (of 14.2%), particularly in women aged 55 -64 years (21.0%), could be explained by earlier detection. Increases in incidence have been recorded in other Western countries in relation to screening (Quinn and Allen, 1995;Garne et al, 1997;Persson et al, 1998;Chu et al, 1996;Rostgaard et al, 2001). In the ES area, no significant increase in incidence was detected. It is likely that an increase in incidence in the ES area should have occurred in the 1970s, as a result of screening.
Incidence trends in the ES and LS areas do not seem to explain the observed differences in mortality rates during the study period. The present results seem to confirm that service screening can reduce breast cancer mortality rates. Table 1 Estimated percentage change (EPC) and 95% confidence intervals (95% CI) of mortality rates (1985 -2000) and incidence rates (1985 -1999) in Early Screening area and in Late Screening area Mortality (1985Mortality ( -2000 Incidence (1985 -1999)