Ductal carcinoma in situ of the breast, a population-based study of epidemiology and pathology

In a population-based series of 2109 women with ductal carcinoma in situ (DCIS) diagnosed in 1995–2000 in New South Wales, Australia, incidence increased by an average of 5.5% a year, mostly between 1995 and 1996 and in women 50–69 years of age. This increase paralleled the increases in mammographic screening. BreastScreen NSW, an organised mammographic screening programme, detected 65% of all DCIS. High-grade lesions were 54% of all lesions and were more likely to be 2+ cm in diameter (OR=2.12, 95%CI 1.46–3.14) than low-grade lesions. In all, 40% of DCIS in women younger than 40 years was 2+ cm in diameter compared with 21% in women 40 years and older. Young age, high grade, mixed architecture and multifocality were significant and independent predictors of 2+ cm DCIS.

The epidemiology of ductal carcinoma in situ (DCIS) has yet to be described adequately. While mammographic screening has undoubtedly caused increasing diagnosis of DCIS (Ernster et al, 1996(Ernster et al, , 2002Levi et al, 1997;Barchielli et al, 1999), studies to date have mainly been small (Levi et al, 1997;Barchielli et al, 1999) or described populations that had no organised screening programme (Choi et al, 1996;Ernster et al, 1996;Zheng et al, 1997). Population-based descriptions of epidemiology and pathology of DCIS in sizeable screened populations are nonexistent.
We describe the epidemiology and pathology of newly diagnosed DCIS in the 2.7 million female population of New South Wales (NSW) Australia in 1995 -2000.

MATERIALS AND METHODS
Data NSW women with a first diagnosis of DCIS in 1995 -2000 and notified to the NSW Central Cancer Registry were eligible for the study; those with a previous or simultaneous (same month) diagnosis of invasive breast cancer were excluded. Ductal carcinoma in situ has been notifiable in NSW since 1993, and by 1997 all but 4% of DCIS cases diagnosed by pathology laboratories were notified to the Cancer Registry (unpublished data).
Two experienced Cancer Registry personnel extracted information on the type of specimen, size, grade, architecture, presence or absence of necrosis and multifocality of DCIS, and clearance and width of the margins from pathology reports (Kricker et al, 1999).
The frequency of mammography was obtained from reports of an organised screening programme, BreastScreen NSW (Estoesta et al, 2000;Productivity Commission, 2002), which began in 1991 and reached a steady state between 1995 and 2000. The numbers of bilateral mammograms reimbursed by the national health insurance scheme, Medicare, were also available for 1995 -99 (http://www.hic.gov.au/providers/health_statistics/statistical_reporting/medicare.htm). These mammograms would include an unknown but not high proportion of mammograms that were primarily diagnostic. BreastScreen and Medicare account for most of the screening mammography in Australia.

Analyses
Incidence rates and 95% confidence intervals (CIs) (Dobson et al, 1991) were calculated in 5-(DCIS) or 10-year (mammography) age groups and age-standardised to the World population. The annual percentage changes in rates were estimated in negative binomial models with terms for age group and year of diagnosis.
Cases were allocated to urban or rural areas using BreastScreen's classification (Estoesta et al, 2000) and to five socioeconomic (SES) groups (Australian Bureau of Statistics, 1998), and variation among them tested in Poisson regression models.
The heterogeneity of DCIS distributions by size and grade among age groups and type of architecture among years of diagnosis was evaluated by standard w 2 tests. Age, grade, architecture, multifocality and presence of necrosis were examined as predictors of size (o2 cm, 2 þ cm) in logistic regression models that included year of diagnosis; the additional effects of urban or rural residence and SES of the women were also examined.

Incidence
In 1995 -2000, 2109 NSW women were notified with DCIS. More than half (54%) were 50 -69 years of age, the target age group for breast cancer screening in NSW, and had the highest incidence (32.3 per 100 000) (Table 1).
The incidence of DCIS was about 25% higher in urban than rural areas of NSW (Po0.001) and increased strongly with increasing socio-economic status in Sydney (Po0.001, Table 1), but not other areas of the State (P ¼ 0.09). BreastScreen NSW detected 65% of incident DCIS, with higher proportions in women older than 50 years.

Screening
BreastScreen and Medicare screened 265.1 per 1000 women 40 years of age and older in 1995 and 284.6 in 1999; all the increase was in BreastScreen. BreastScreen screened twice as many women 50 -69 years of age (269.2 per 1000) as women 40 -49 (122.4) and 70 þ years (120.1) ( Figure 2); Medicare screening rates in the two younger age groups were nearly equal. There was little difference in the screening rates between urban (189.0 per 1000) and rural (186.6) areas.
The numbers of DCIS detected were strongly correlated with the numbers of women screened by BreastScreen in 1995 -2000 (R ¼ 0.83, Po0.001). There was a weaker correlation between Medicare reimbursed mammograms and DCIS not detected by BreastScreen in 1995 -1999 (R ¼ 0.64, P ¼ 0.002).

Pathology information
Pathology reports were available for all but two cases of DCIS, mainly from excision or re-excision specimens (73%) and mastectomies (24%). Size was reported in 76%, grade in 89% (both in 69%), presence or absence of necrosis in 63%, architecture  in 82% and clearance of the margins in 80%; 37% had all these items. The width of margins was reported in only 41%. All these items except the width of margins were appreciably more complete in 1998 -2000 than in 1995 -97.
Age, grade, architecture and multifocality were significant independent predictors of DCIS 2 þ cm in size in a logistic regression model (Table 3). Two-fold higher odds of having 2 þ cm DCIS appeared to lie with a young age (20 -39 years), with high grade and with multifocal lesions. The excess risks of 2 þ cm DCIS also lay with mixed types of architecture (OR 1.5) compared to comedocarcinoma and diagnosis in 1996 (OR 1.7), while a report that did not mention necrosis was associated with lower odds (OR 0.7) of a larger DCIS. Otherwise, there was no evident trend across these variables. Neither place of residence (urban or rural) nor SES was significantly predictive of 2 þ cm DCIS or appreciably affected the above odds ratios when added to the model.

DISCUSSION
The main strengths of this study are its population base, detail and recency. Population-based registration of DCIS in NSW began 2 years before the first year of our study, and the study's first year coincided with that of complete population coverage by BreastScreen. Its main weakness is lack of linkage between the data sources (BreastScreen, Medicare and the Cancer Registry), without which we cannot fully describe the contribution of screening to the occurrence and outcomes of DCIS.
Other countries have observed increasing incidence of DCIS, sometimes three-to four-fold, with increasing mammographic screening (Ernster et al, 1996;Levi et al, 1997;Zheng et al, 1997;Barchielli et al, 1999;Blanks et al, 2000). This increase has stopped in NSW and we would expect further increases only with growth in BreastScreen participation beyond 53 -54% in 199853 -54% in -200053 -54% in (Estoesta et al, 2000Productivity Commission, 2002) or resurgence in mammographic screening reimbursed by Medicare.
We observed a higher proportion of high-grade lesions (54%) than in Sweden (43%) and Switzerland (46%) in the early 1990s, but similar to that, 55%, in an Australia-wide sample survey (Levi et al, 1997;Wärnberg et al, 1999;Shugg et al, 2002). This apparent difference between Australia and these European countries may be real and due to the extent of high-grade disease, or caused by differences in the reporting of grade. It could reflect, too, differences in mammography rates since high-grade DCIS is said to show abnormal mammographic features more frequently than low grade (Evans et al, 2001).
We found that age, high-grade lesions, mixed architecture and multifocality significantly and independently predicted DCIS larger than 2 cm diameter. That women 20 -39 years of age were more likely to have 2 þ cm diameter DCIS (average 28 mm in multivariate models in this study) than older women (average 18 mm) is probably due to their lower mammography rate. That high-grade lesions were larger probably reflects a correlation with higher growth rate (1.8 mm per year low grade, 4.2 mm   Thomson et al, 2001), particularly since they may also be more readily detectable mammographically (Evans et al, 2001). The Van Nuys prognostic index has shown a parallel increase of size with grade although patients treated with mastectomy, and thus probably the larger lesions, were excluded (Silverstein, 2003). Another series showed a step down from 20 mm diameter poorly differentiated DCIS to 15 mm for all other grades (Solin et al, 1991). Our study appears to give the first population-based estimates: the average size increased steadily from 16 mm for low grade to 20 mm for intermediate and 27 mm for high-grade DCIS in the multivariate models of Table 3. Variation in growth rate might also underlie the significant, independent association of architecture with tumour size. The pathology reports from 1998 to 2000 reported DCIS substantially more completely than did earlier Australian reports (Kricker et al, 1999;Giles et al, 2001;Shugg et al, 2002). The Australian Cancer Network addressed pathology reporting of DCIS with extensive consultation among pathologists in the mid-1990s and published recommendations in 1997 (Australian Cancer Network Working Party, 1997(Australian Cancer Network Working Party, 2001. Their adoption by pathologists may explain the more complete reporting we observed.
Increasing incidence of DCIS is an outcome of mammographic screening for breast cancer. The high proportion of high-grade lesions we have observed suggests that its detection could contribute to reducing breast cancer mortality. If it does not, the high frequency of DCIS in association with screening may be source of unnecessary morbidity and cost. More research is needed into the costs and benefits of detection of DCIS in breast cancer screening.