Estimating risk of breast cancer from population incidence affected by widespread mammographic screening.

Abstract

To estimate the absolute risk of breast cancer in women, allowing for the effect on incidence of the introduction of widespread mammographic screening. Annual breast cancer incidences were compared with numbers of annual mammograms in the population for 1980-96 to identify periods most likely to be affected by screening. Age specific breast cancer incidences 1972-96 were modelled by Poisson regression with an age, period, and cohort analysis. The 1996 age specific incidence was recalculated with the stable period effect 1972-89, and the age and cohort effects. Age specific incidence was converted to cumulative risk of breast cancer to age 79. Population based data from all women in New South Wales (NSW), Australia. Breast cancer incidence in women 1972-96 obtained from the NSW Central Cancer Registry and female populations derived from successive censuses. Mammographic data from private sector mammograms (1985-96), and the mammographic screening service (1988-96) for NSW women. Introduction of population mammographic screening. Recorded age specific incidence and absolute risk of breast cancer to age 79 was compared with underlying incidence and cumulative absolute risk, adjusted for recent period effects, most likely due to mammographic screening in the population. The age, period, and cohort model showed an increasing effect for birth cohorts 1910-44 then a plateau, and prominent period effects in 1991 and 1994-6. Increased incidence of breast cancer coincided with an increase in mammographic examinations in the private sector (1991), and prevalent rounds of mammographic screening in the population (1994-6) after introduction of a statewide mammographic screening service. Recorded incidence produced a breast cancer risk to age 79 of 9.9% (1 in 10) for 1996, whereas estimation of underlying incidence yielded a risk of 8.5% (1 in 12). The introduction of mammographic screening in a population inflates the incidence of breast cancer because of diagnosis of prevalent cases. For the purpose of public and clinical communication, it is more reasonable and responsible to adjust for period effects (due to screening) rather than produce risk estimates based on recorded incidence, which may show an alarming increase in risk of breast cancer over a short period.