This large, retrospective, population-based study of women aged 50–74 has found that annual screening mammography was associated with a higher proportion of screen-detected cancers than was biennial screening. The distribution of cancers and prognostic profiles translated into small absolute differences in modelled 5- and 10-year breast cancer-specific survival in a hypothetical cohort of 100
000 screens performed either annually or biennially (0.6 and 1.2%, respectively), in favour of the annually screened group. In keeping with this, there was no difference in observed survival in women diagnosed with ipsilateral, invasive breast cancer undergoing screening mammography during the time of an annual or biennial screening policy.
The predicted differences in 10-year breast cancer survival were similar to the findings of the only randomised trial addressing this issue, the United Kingdom Co-ordinating Committee on Cancer Research Randomised Trial performed (Breast Screening Frequency Trial Group, 2002
). That study randomised women, aged 50–62 years, with a normal prevalent screen in the UK National Health Services (NHS) Breast Screening Programme, to a conventional incident screen after an interval of 3 years (control arm) or to three annual screenings (study arm).
The BSFTG randomised trial found that the incidence of breast cancers was 2.44 per thousand per annum in the study arm and 2.15 in the control arm, similar to the estimates from our study. Also, 71% of the cancers were screen-detected in the annual recall (study) arm compared to 50% in the 3-year recall (control) arm, similar to the rates, 76% in the annual and 58% in biennial screening cohorts found in this study. As expected, the shorter screening interval was associated with a higher proportion of screen-detected cancers, compared to interval cancers, but this translated to only a small absolute difference in the estimated 10-year breast cancer-specific survival: approximately 2% in the randomised trial and 1% in the current study.
In a meta-analysis of data from the randomised trials, there was no difference in the reduction in breast cancer mortality among women aged 50–74 with planned screening intervals of less than 18 months compared to those with longer intervals of 18–33 months (Kerlikowske et al, 1995
Computer modelling simulations suggest that breast cancer mortality reductions improve with shorter screening intervals, with improvements as much as from 26%, for mammography performed every 5 years, to 66%, annual mammography (Szeto and Devlin, 1996
; Michaelson et al, 1999
; Fett, 2001
). In contrast, Jansen et al
used modelling to show that shorter screening intervals may be associated with greater detection of ‘excess' tumours that would not have been detected before the subject died of other causes, and that the overall benefit to the population with respect to breast cancer mortality was greater when a larger proportion of the population were screened at a lower frequency compared to a smaller proportion screened more frequently (Jansen and Zoetelief, 1997
Retrospective studies comparing interval and screen-detected cancers have shown that interval cancers are similar to cancers in unscreened women, and that screen-detected cancers generally have better prognostic indicators, with smaller tumours, fewer node-positive tumours, and a larger proportion of in situ
tumours (Burrell et al, 1996
; Porter et al, 1999
). The current study confirms these observations, but of note is the similarity of the prognostic profile of interval cancers for those screened annually and that of interval cancers in those screened biennially, with the same being true for screen-detected cancers.
Retrospective cohort studies that examined prognostic differences of tumours detected in women who self-selected different screening frequencies have generally shown that those screened less frequently have tumours with worse prognostic profiles, including larger tumour size, more lymph node metastases, and fewer cases of isolated in-situ disease (Gabriel et al, 1997
; Field et al, 1998
; Carlson et al, 1999
; Hunt et al, 1999
). It has been shown that the higher absolute rate of false-positive screens increased with more frequent screening (Elmore et al, 1998
A limitation of the current study is that, during the study interval (1995–2002), there could have been improvements in the technical aspects of imaging or in radiologist experience that may have affected detection rates (Kan et al, 2000
; Feig, 2002
; Haus, 2002
; Taplin et al, 2002
), although SMPBC data suggest that improved detection is unlikely to explain our observations. Within the SMPBC, women, aged 40–49, were recalled for screening annually both before and after 1997 and comparison of these periods showed little difference, with the cancer detection rate increasing only 0.1 per 1000 for both first and subsequent screens after 1997 (SMPBC unpublished data, December 2003). Another limitation was the incomplete enforcement of the change in screening policy with some women not following the re-screening recommendations in each period. This necessitated a wider range of duration between screens in the era after 1997 to create comparably sized groups for analysis purposes.
In spite of these limitations, this study provides useful information about different screening intervals based on actual data from a large, population-based screening programme that instituted a policy change, in the modern era of two-view, film-screen mammography. Complete and accurate outcome data were obtained through direct linkage to the provincial cancer registry. The two comparison groups were selected to minimise selection bias. The predicted survival was based on a prognostic model that used cases over the whole period to avoid the survival effect of changes in treatment between the different periods. Thus, the predicted survival was a ‘blended' rate of the survival in each period and was applied uniformly to everyone in the study regardless of her date of diagnosis.
In conclusion, this study has shown that annual screening mammography minimally improved the estimated breast cancer survival rates for women aged 50–74 years, as compared to biennial screening at 5 and 10 years, with an absolute projected survival difference of 0.6% at 5 years and 1.2% at 10 years for those women diagnosed with breast cancer.