Mammography screening and incidence of ductal carcinoma in situ of the breast in Italy: an age-period-cohort analysis

Abstract

Background: The increasing incidence of ductal carcinoma in situ (DCIS) of the breast is attributed to mammography screening, but the trend has so far been evaluated only descriptively.

Methods: We report an age-period-cohort modelling analysis of the incidence trend (1992-2017) observed among screening-aged women living in a region of northern Italy, where a mammography screening programme was implemented in 1996-98 (age 50-69 years) and 2010-14 (age 45-49 and 70-74 years).

Results: The incidence of DCIS increased by an annual average of 9.1% (age 60-64 years) to 14.9% (age 70-74 years). The incidence peak followed a complex age-time pattern indicating an interaction between age and period, which suggested a cohort effect explained by the screening programme. In the age-period-cohort analysis, the birth cohort had a 2-fold effect. The nonlinear effect consisted of an increase in incidence for the generations of 1933-42 and 1943-52, targeted by screening since 1997, and of a second increase for the cohort of 1963-72, first invited in 2010. Taking into account the early excess incidence due to the introduction of the screening programme, the linear effect consisted of an annual 4.0% increase in the risk of DCIS for all successive birth cohorts or calendar periods, which was only partially attributable to the programme.

Conclusions: The increase in incidence rates resulted from an increased detection of DCIS through the screening programme and from an uptrend in the risk of diagnosis that can be attributed either to long-term changes in diagnostic scrutiny independent of the programme or to an increased exposure to risk factors.

Keywords: age-period-cohort analysis; breast neoplasms; ductal carcinoma in situ; incidence; mammography; mass screening.

Figure 1.

Lexis diagram showing the 5-year calendar period-specific (6 years for the period 2012–17), the 5-year age groups, and the 10-year birth cohorts. Highlighted in yellow are the birth cohorts invited to the screening programme in specific age groups and calendar periods. Romagna (northern Italy), 1992–2017.

Figure 2.

Observed and smoothed (LOWESS) curves of annual incidence rates of ductal carcinoma in situ of the breast per 100 000 women, by 5-year age groups. Romagna (northern Italy), 1992–2017.

Figure 3.

Incidence rates of ductal carcinoma in situ of the breast per 100 000 women and number of cases, by 5-year age groups and 5-year calendar periods (6 years for the period 2012–17). The cells highlighted in orange contain the data for women aged 50–69 years, invited to the screening programme since 1997. The cells highlighted in green contain the data for women aged 45–49 and 70–74 years, invited since 2010. The cell highlighted in blue contains the data for the cohort of women invited once in 1997, when they were in the age range 50–69 years. The remaining cells contain the data for women never invited. Romagna (northern Italy), 1992–2017.

Figure 4.

Plots of 5-year age group-specific incidence rates of ductal carcinoma in situ of the breast per 100 000 women, on a semilogarithmic scale, by 10-year birth cohorts. Romagna (northern Italy), 1992–2017.

Figure 5.

Cohort effect estimates from the age-period-cohort model fitted on incidence rates of ductal carcinoma in situ of the breast per 100 000 women. The cohort effect estimates (rate ratios) were obtained taking the birth cohort of 1948–57 as a reference (red line). The blue line (with shaded 95% confidence interval) indicates the pooled linear and nonlinear cohort effects. The orange line indicates the nonlinear cohort effect alone (with shaded 95% confidence interval). Romagna (northern Italy), 1992–2017.