NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Virnig BA, Shamliyan T, Tuttle TM, et al. Diagnosis and Management of Ductal Carcinoma in Situ (DCIS). Rockville (MD): Agency for Healthcare Research and Quality (US); 2009 Sep. (Evidence Reports/Technology Assessments, No. 185.)

  • This publication is provided for historical reference only and the information may be out of date.

This publication is provided for historical reference only and the information may be out of date.

Cover of Diagnosis and Management of Ductal Carcinoma in Situ (DCIS)

Diagnosis and Management of Ductal Carcinoma in Situ (DCIS).

Show details

3Results

This review addresses four related questions about DCIS. The first question addresses DCIS incidence and detection. The second, DCIS diagnostic evaluation with MRI and the utility of sentinel lymph node biopsy. The third addresses nontreatment factors associated with DCIS outcomes, and the final question addresses the impact of treatment on DCIS outcomes. Figure 5 outlines the results of the literature review process, the articles identified, and those ultimately deemed eligible.

Figure 5. Study Flow.

Figure 5

Study Flow.

Question 1. What are the incidence and prevalence of DCIS and its specific pathologic subtypes, and how are incidence and prevalence influenced by mode of detection, population characteristics, and other risk factors?

The incidence of DCIS is gaining attention as it is increasing from a relatively rare finding in the 1970s to a finding representing up to 25 percent of all breast cancers by 2004. In this chapter we review factors related to the incidence of DCIS and, to the extent possible, place them in the context of invasive breast cancer.

We identified 63 publications from population based studies that reported the incidence of DCIS;8,17,44–104 36 studies were conducted in the United States (Appendix Table F1).17,44–46,48–50,52,56,58–60,66,68,70–75,77,80–82,85,87–92,95,97,99,101,103 We identified 29 studies (Appendix Table F2) that examined risk factors for DCIS.80,99,105–112 88,113 92,114,115 68,116–128 Eight population-based mammography trials evaluated the effect of mammography on DCIS and invasive breast cancer incidence.129–136

Incidence of DCIS per 100,000 Standardized Female Population

Population-based cancer registries offer some of the strongest evidence for changing incidence of DCIS. We identified 11 studies analyzing the Surveillance Epidemiology and End Results (SEER) database and state cancer registries to report incidence of DCIS per 100,000 standard U.S. female populations (Appendix Table F3).17,56,59,73,74,77,80,82,90,91,95 Among foreign studies, 12 retrospective cohorts,53–55,61,62,67,69,76,78,83,86,102 and two RCTs reported incidence rates per 100,000 female population (Appendix Table F4).51,57

Incidence over time. Regardless of source, the incidence of DCIS has increased dramatically since the early 1970s. The National Cancer Institute (NCI) report SEER Cancer Statistics Review 1975–2004 estimated the incidence of DCIS in 2004 to be 32.5 per 100,000 women. While considerably higher than the 5.8 per 100,000 in 1975, the rate is considerably less than the invasive breast cancer incidence estimated to be 124.3 per 100,000 in 2004. These same trends are reported in numerous studies using the SEER registries as a whole as well as single registries or groups of registries.17,59,77,82,90,95 The incidence, however, was not stable across all DCIS subtypes. DCIS with comedo necrosis, a particularly aggressive subtype of DCIS, has not increased, while the increase in incidence of noncomedo DCIS increased 15–22 times.82

While other countries have also reported increases in DCIS, no country currently reports rates as high as those observed in the United States. Age adjusted annual incidence of DCIS in the 1990s was the lowest in Switzerland (3.95 per 100,000) and Italy (6 per 100,000), with the highest incidence in The Netherlands (11 per 100,000) (Figure 6 and Appendix Tables F45).51,61,67,69,76,78,83

Figure 6. Time trend in age adjusted annual incidence of DCIS per 100,000 females (results from individual studies).

Figure 6

Time trend in age adjusted annual incidence of DCIS per 100,000 females (results from individual studies).

A series of autopsy studies examined the prevalence of undiagnosed DCIS among women who died of reasons other than breast cancer. These studies were, without exception, conducted prior to routine use of mammography and pointed to prevalence of unrecognized DCIS ranging from less than 1 percent to 14.3 percent. These same studies found smaller amounts of unrecongized breast cancer (less than 2 percent when reported) (Table 1).

Table 1. Prevalence of occult DCIS in autopsy studies.

Table 1

Prevalence of occult DCIS in autopsy studies.

Risk Factors for DCIS

In general, the risk factors that are explored for DCIS are the same factors that are associated with invasive breast cancer. These risk factors are grouped into several broad categories: (1) demographic factors, (2) reproductive factors, (3) biological risk factors such as family history, (4) behavioral risk factors, and (5) screening using mammography. A sixth category is chemoprevention and detection of DCIS for high risk women.

Demographic factors.

Age-specific incidence of DCIS. The incidence of DCIS, like invasive breast cancer, is strongly related to age. Incidence of DCIS in the United States per 100,000 women is extremely uncommon prior to age 35–39 (2.5 per 100,000 for women ages 30–34). After that, the incidence rises steadily to a peak of 96.7 per 100,000 at ages 65–69 and then declines, slowly until age 79 and steeply after that.77,82,91,95 In contrast, invasive breast cancer peaks at age 75–79 with incidence of 453.1 per 100,000 women (Figure 7). At no age is DCIS more common than invasive breast cancer. Between the ages of 40 and 64, between 21 and 22.8 percent of all breast cancers are DCIS. Prior to age 40 and after age 64 the proportion of breast cancers that are DCIs drops to as low as 9 percent. Studies of change in incidence of DCIS over time point to increases in all age groups but are the greatest among women older than 50 years.77,82,95

Figure 7. Incidence of DCIS and invasive breast cancer by age (2002–2006).

Figure 7

Incidence of DCIS and invasive breast cancer by age (2002–2006).

Race. Several studies report the incidence of DCIS by race or ethnicity. The overall age-adjusted incidence rates per 100,000 population were the same in whites when compared to nonwhites.117 However, when examining racial groups more closely, the age adjusted incidence of DCIS was the highest among Caucasian women (Appendix Table F6) followed by African American and Asian-Pacific Islanders (Figure 8).73,80 Hispanic women had the lowest age adjusted incidence of DCIS. Consistent with these registry-based findings, five studies examined the association between race and DCIS and with one exception reported African Americans had lower incidence of DCIS than whites. The studies did not find any remarkable differences in DCIS between white and Asian women (Appendix Table F7).80,88,115,117,123 It is important to note the lower rates of DCIS for African American, Asian, and Hispanic women, coupled with lower rates of invasive cancer. Thus, the evidence does not suggest that lower rates of DCIS in nonwhites should be viewed as indicating a failure to diagnose breast cancer early but could be related to lower underlying risk of breast cancer.

Figure 8. Age-adjusted rates of DCIS and invasive breast cancer, SEER 2002–2006, by race.

Figure 8

Age-adjusted rates of DCIS and invasive breast cancer, SEER 2002–2006, by race.

Urban/rural. One study used the SEER data to examine the change in DCIS incidence for urban and rural women.74 That study found that prior to 1973 there were no urban/rural differences between urban and rural-dwelling women. After 1973 the incidence of DCIS rose in both groups but rose more steeply in urban women than in rural women. The study did not offer comparable estimates of the incidence of invasive cancer or total breast cancer (DCIS plus invasive) to provide context. Similar effects of residence were found in Australia, where urban-dwelling women were diagnosed more often with DCIS (9 per 100,000) than women from rural areas (7.1 per 100,000, 95 percent CI 6.3; 7.8).76

Education. A single study examined the role of education and found that less educated women (<high school) had greater cumulative incidence of DCIS from January 1997 to December 2001 (7.3 percent) compared to women with higher education (4.5 percent).85

Income. A single Australian study linked DCIS incidence to socioeconomic status and found that the cumulative incidence of DCIS was the lowest in women of the lowest socio-economic status (7.2 per 100,000) compared to women with the highest status (11.2 per 100,000).76

Reproductive factors.

Age at menarche. Three studies examined the association between odds of DCIS and age at menarche.109,116,120 While there was a slight trend toward decreased odds of DCIS associated with older age at menarche, no study found a statistically significant association (Figure 9).117

Figure 9. Association between age at menarche and DCIS.

Figure 9

Association between age at menarche and DCIS.

Age at menopause. Age at menopause is challenging to examine in the context of DCIS because the risk of DCIS increases with age, particularly around the age of menopause (45–60). Thus, it can be challenging to separate the effects of aging with the hormonal changes associated with menopause. A study based on the New York Tumor Registry found significantly increased risk of DCIS for peri- and post-menopausal women compared to pre-menopausal women (Figure 10). Only the study based on the Connecticut Tumor Registry found a significant association between age at menopause and DCIS. That study found the women who were over age 55 at menopause had increased risk of DCIS compared to women who were less than 45 at menopause.120 No other study found a significant positive association between increased odds of DCIS and older age at menopause. The Iowa Women’s Health Study found a slight, nonsignificant increase in the relative risk of DCIS among women undergoing natural menopause versus surgical menopause (RR 1.19, 95 percent CI: 0.87–1.63).109 The Connecticut study also reported that for each year menopause is delayed, relative odds of DCIS rise by 2 percent.120

Figure 10. Association between menopause and DCIS.

Figure 10

Association between menopause and DCIS.

Hormone replacement therapy. The association between hor mone replacement therapy (HRT) and DCIS was examined in five observational studies (Appendix Table F8).68,108,109,112,120 Neither the Iowa Women’s Health Study109 nor studies based on the Breast Cancer Surveillance Consortium database or state cancer registries found an association between ever (versus never) use of HRT and increased risk of DCIS.112,120 A large prospective cohort study in the United Kingdom based on the National Health Service Central Registers108 found a 56 percent increased risk of DCIS in current users of HRT compared to never users (Figure 11). Two studies (the Iowa Women’s Health Study and the Breast Cancer Screening Consortium) found that the increased risk of DCIS with HRT varied with duration of use. Current users of hormone replacement therapy for less than 5 years compared to never users had significantly less risk of DCIS (pooled relative risk [RR] 0.78, 95 percent CI 0.63; 0.96).109,112 Studies of current users of HRT for more than 5 years found the opposite association, with greater risk of DCIS compared to never users (pooled RR 1.41, 95 percent CI 1.24; 1.59) (Figure 12).109,112 The Iowa Women’s Health Study found no increased risk of DCIS among prior users of HRT compared with never users.109 In contrast, a case control study based on Wisconsin’s Cancer Registry reported increased odds of DCIS among past users compared to never users.68 The United Kingdom study also found an increased risk of DCIS among past users compared to never users.108

Figure 11. Association between ever use of hormone replacement therapy and DCIS.

Figure 11

Association between ever use of hormone replacement therapy and DCIS.

Figure 12. Association between hormone replacement therapy and DCIS.

Figure 12

Association between hormone replacement therapy and DCIS.

The increased risk of invasive breast cancer associated with HRT is well established and reported in both observational and randomized studies. The Women’s Health Initiative, a large randomized trial of HRT and breast cancer risk, found no increased risk of DCIS associated with HRT.137,138 The large Million Women Study cohort, failed to comment on whether they observed any increase in DCIS associated with HRT use.

Oral contraceptive use. The association between oral contraceptives and DCIS was examined in five studies (Appendix Table F8).68,118,120,122,126 Women who had ever used oral contraceptives,68,120,122,126 were current users, or who used contraceptive sometime in the past126 had the same odds of DCIS as never users (Figure 13). Two studies failed to find a significant association between the duration of oral contraceptive use and DCIS incidence (Figure 14).122,126 The association between ever use of oral contraceptives and DCIS in women with and without family history, and post- and pre-menopausal women was not significant in the case control study based on the Connecticut Tumor Registry (Figure 15).126 The Connecticut Tumor Registry study126 found no significant differences in odds of DCIS by type of contraceptives, estrogen dose (low or high), or progestin types when compared to never users. Studies of whether age at oral contraceptive use influenced risk did not point to age being an important effect modifier (Figure 16).

Figure 13. Association between oral contraceptives and DCIS.

Figure 13

Association between oral contraceptives and DCIS.

Figure 14. Association between duration of oral contraceptive use and DCIS.

Figure 14

Association between duration of oral contraceptive use and DCIS.

Figure 15. Association between ever use of oral contraceptives and DCIS in subgroups by family history of breast cancer and menopausal status (multivariate adjusted odds ratio from the study based on the Connecticut Tumor Registry).

Figure 15

Association between ever use of oral contraceptives and DCIS in subgroups by family history of breast cancer and menopausal status (multivariate adjusted odds ratio from the study based on the Connecticut Tumor Registry).

Figure 16. Association between oral contraceptive use and DCIS by starting age.

Figure 16

Association between oral contraceptive use and DCIS by starting age.

Parity. The association between parity and DCIS was examined in seven studies (Appendix Table F9).68,109,111,116,120,123,128 The studies that examined the association between DCIS and age at first live birth compared to less than 20 years found a significant increase in the risk of DCIS among those who had their first child between 20 and 29 years (pooled RR 1.43, 95 percent CI 1.07; 1.91) and more than 30 years of age (pooled RR 1.46, 95 percent CI 1.27; 1.67) but not among other age categories (Figure 17).68,109,120,123 Women who had their first live birth between 25–34 years of age had increased risk of DCIS compared to those 20–24 years of age, according to the Danish Breast Cancer Cooperative Group registry (Figure 18).111 One case control study from the Rapid Case Ascertainment Shared Resource at the Yale Cancer Center reported a borderline significant positive association between older age at the first birth and DCIS (odds ratio [OR] 1.02, 95 percent CI 1; 1.05).120 The University of California San Francisco Mobile Mammography Screening Program found that nulliparous women or women older than 30 years at birth of their first child had 130 percent greater odds of DCIS than women who had children prior to age 30.116 The Danish cohort also found that women who had the first live birth after age 30 had an increased risk of larger tumors and comedo type DCIS (Figure 19).111

Figure 17. Association between DCIS and age at first live birth compared to less than 20 years.

Figure 17

Association between DCIS and age at first live birth compared to less than 20 years.

Figure 18. Association between DCIS and age at first live birth among different age categories.

Figure 18

Association between DCIS and age at first live birth among different age categories.

Figure 19. Association between types of DCIS and age at first live birth compared to 20–24 years (Danish Breast Cancer Registry).

Figure 19

Association between types of DCIS and age at first live birth compared to 20–24 years (Danish Breast Cancer Registry).

The association between number of births and DCIS was examined in six studies (Appendix Table F10).109,111,116,120,123,128 Women with four or more children had a 38 percent decreased risk of DCIS compared with women with one child (pooled RR 0.62, 95 percent CI 0.43; 0.90).111,123 Similar decreased risk associated with having three or more children relative to one child or no children was reported by a large Swedish registry based study.128 A case control study120 found a significant dose response association between greater number of births and reduced odds of DCIS; however, a large Danish Breast Cancer Cooperative Group cohort did not show such protective effect of parity (Figure 20).111

Figure 20. Association between parity and DCIS.

Figure 20

Association between parity and DCIS.

Biological risk factors.

Breast density. Premenopausal women with heterogeneous or extreme breast density had higher risk of developing DCIS than women with scattered density.99 Postmenopausal women with heterogeneous breast density had a higher risk of DCIS (RR 1.41), while women with fatty breasts developed DCIS less often (RR 0.58) when compared to women with scattered breasts (Figure 21).99 A nested case control study also found increased odds of DCIS among women with higher than 50 percent versus lower than 10 percent mean breast density (OR 2.86, 95 percent CI 1.38; 5.94) (Figure 22).92 Women with a mean breast density of >45 cm2 also had greater odds of DCIS than women with a low breast density <15 cm2 (OR 2.59, 95 percent CI 1.39; 4.82).92

Figure 21. Association between breast density, previous history of breast biopsy or surgery, and DCIS.

Figure 21

Association between breast density, previous history of breast biopsy or surgery, and DCIS.

Figure 22. Adjusted odds ratios of DCIS by mammographic breast density (results from the Multiethnic cohort.

Figure 22

Adjusted odds ratios of DCIS by mammographic breast density (results from the Multiethnic cohort.

Body composition. Three studies examined the association between body mass composition and DCIS (Appendix Table F11).109,116,123 One case-control study based on the SEER database reported that the odds of DCIS were greater in women with body mass index (BMI) <22kg/m2 (Figure 23).123 The Iowa Women’s Health study did not find greater risk of DCIS in women with BMI <24 compared to overweight or obese women.109 Women with BMI >25 among women 30–49 years old but not among those older than 50 years had increased odds of DCIS.116The Iowa Women’s Health Study also failed to find an association between waist-to-hip ratio, a measure of abdominal adiposity, and DCIS incidence.109 Kerlikowske found increased odds of DCIS among women with BMI greater than 25 who were between 30 and 49 years but not for women older than 50 years.116 A single study found that heavily obese (BMI ≥35.0 kg/m 2) postmenopausal women not taking hormone replacement therapy had increased odds of DCIS (OR 1.46, 95 percent CI 1.14; 1.87) relative to normal weight women after adjustment to race, ethnicity, age, mammography use, and registry.139

Figure 23. Association between body composition and DCIS.

Figure 23

Association between body composition and DCIS.

Family history. Several studies reported that women with a family history of breast cancer or a first degree relative with breast cancer had similarly increased odds of DCIS compared to women without a positive family history (pooled OR 1.97, 95 percent CI 1.10, 3.52) (Figure 24).68,85,116,120 One study found that the increased risk associated with having a sister with breast cancer was greater for younger women than older women (OR 3.74 versus 2.1).

Figure 24. Family history of breast or ovarian breast cancer and DCIS.

Figure 24

Family history of breast or ovarian breast cancer and DCIS.

Several European surveillance trials reported DCIS incidence among BRCA1/2 gene mutation carriers and women with high familial risk (Appendix Table F12).140–147 Annual DCIS incidence varied from 0.1–1.5 percent in the Netherlands145–147 to 0.9 percent in Canada.142 Other studies reported intermediate rates: 0.2–0.6 percent in Norway140,141 and 0.2–0.4 percent in the United Kingdom.143,144 A U.S. study of similarly high risk women found the cumulative crude incidence of DCIS over 7 years to be 9.1 percent (95 percent CI 2.3; 30) (Appendix Table F13).148 A cohort of 1,198 women followed for 3 years in the Netherlands147 reported higher rates of DCIS among BRCA1/2 gene mutation carriers (0.4 percent) and among those with estimated risk of breast cancer more than 25 percent (0.6 percent, 95 percent CI 0.2; 1.7).

A study based on the Connecticut Tumor Registry did not observe a significant association between family history of ovarian cancer and DCIS.125

The association between DCIS and common variants on chromosome 5p12 was investigated in a multinational case control study pooling individual patient data from 6,145 cases and 33,016 controls in several countries (Appendix Table F14).127 Women with a single nucleotide polymorphisms rs4415084 and rs10941679127 had increased odds of DCIS (Figure 25).127

Figure 25. Association between DCIS and common variants on chromosome 5p12 confer susceptibility to estrogen receptor-positive breast cancer (odds ratios from the multinational case-control study, adjusted to age and other variables).

Figure 25

Association between DCIS and common variants on chromosome 5p12 confer susceptibility to estrogen receptor-positive breast cancer (odds ratios from the multinational case-control study, adjusted to age and other variables).

Blood levels of lipids, proteins, sex hormones, and mitogenes. The association between DCIS and blood levels of biologically active substances was examined in three studies (Appendix Table F15).114,119,121 The New York University Women’s Health Study did not identify a significant association between sex hormones and odds of DCIS (Figure 26).114 One case control study reported a significant association between balance of mitogenes and odds of DCIS.121 Women at high risk of cancerogenesis defined as higher tertile of insulin-like growth factor-I and the lowest tertile of insulin-like growth factor binding protein-3 had increased odds of DCIS (OR 3.7, 95 percent CI 1.1; 12.2) (Figure 26).121 One hospital-based case control study found no association between serum cholesterol and odds of DCIS.119 The same study reported a dose response increase in odds of DCIS among those with higher albumin levels.119

Figure 26. Age adjusted odds ratio of DCIS among categories of sex hormones (from the New York University Women’s Health Study).

Figure 26

Age adjusted odds ratio of DCIS among categories of sex hormones (from the New York University Women’s Health Study).

Benign breast conditions. The association between DCIS and previous breast biopsy or surgery was examined in six studies (Appendix Table F16).68,92,99,116,120,123 Previous breast surgery was not associated with increased odds of DCIS (Figure 21).116 Two cancer registry based case control studies120 and an analysis based on the SEER database123 reported odds of DCIS in women with previous breast biopsies compared with women with no history of breast biopsy (pooled odds ratio 2.7, 95 percent CI 1.4; 5.1, I2 79.4 percent).120,123 Women previously diagnosed with benign breast disease had increased odds of DCIS by 88 percent (OR 1.88, 95 percent CI 1.32; 2.68).68

Behavioral risk factors.

Alcohol. Three studies examined the association between DICS and alcohol intake (Appendix Table F17).68,109,120 A case control study found a significant increase in the odds of DCIS among women with 39–90g of alcohol/week or ≥91g/week compared to nondrinkers.68 Two other studies, one case control120 and a prospective cohort,109 did not find a significant association between ever versus never drinkers or among those who consume more or less than 4g/day compared to never drinkers (Figure 27).

Figure 27. Association between alcohol and dietary factors and DCIS.

Figure 27

Association between alcohol and dietary factors and DCIS.

Dietary beta carotene. One case control study examined the association between dietary beta carotene intake and DCIS (Appendix Table F17).68 Women with the highest intake of beta carotene (>258 kIU) had lower odds of DCIS compared to those with the lowest intake (<760 kIU) (OR 0.54, 95 percent CI 0.35; 0.84) (Figure 27).

Smoking. One case control study examined the asso ciation between DCIS and smoking and did not find differences in odds of DCIS among ever versus never smokers (Appendix Table F17).120

Physical activity. One case control study, based on the Cancer Surveillance Program and the Women’s Contraceptive and Reproductive Experiences Study,124 examined the association between DCIS and physical activity (Appendix Table F17). Across all age categories, women who exercised more than 4 hours per week had lower odds of DCIS than women who exercised less (Figure 28).124 The association between physical activity and DCIS was strong and consistent among women with lifetime activity of at least 1 hour per week or 3–32 MET hours/week compared to none (Figure 28).124 Physically active women had a 34–47 percent reduction in adjusted odds of DCIS (OR 0.65, 95 percent CI 0.48; 0.9) for lifetime physical activity compared to sedentary life styles.124 The strongest protective effect was seen among currently active women (10 years before the study) (Figure 28). Women who exercised more than 4 hours per week within 10 years before the study had a 48 percent reduction in their odds of DCIS (OR 0.52, 95 percent CI 0.33; 0.8).124

Figure 28. Association between physical activity and DCIS (adjusted odds ratios from the Cancer Surveillance Program and the Women’s Contraceptive and Reproductive Experiences Study).

Figure 28

Association between physical activity and DCIS (adjusted odds ratios from the Cancer Surveillance Program and the Women’s Contraceptive and Reproductive Experiences Study).

Nonsteroidal anti-inflammatory agents. The Iowa Women’s Health Study cohort examined the association between nonsteroidal anti-inflammatory agents and the risk of DCIS (Appendix Table F18).110 The multivariate adjusted relative risk of DCIS was significantly lower among frequent aspirin users compared to nonusers (Figure 29). Surprisingly, the association was not observed for other nonsteroidal anti-inflammatory agents (e.g., ibuprofen).

Figure 29. Multivariate adjusted relative risk of DCIS in association with aspirin and nonsteroidal anti-inflammatory agents (results from the Iowa Women’s Health Cohort Study).

Figure 29

Multivariate adjusted relative risk of DCIS in association with aspirin and nonsteroidal anti-inflammatory agents (results from the Iowa Women’s Health Cohort Study).

Screening using mammography.

Screening. Many researchers and policymakers alike have questioned whether the recongized increase in DCIS incidence is due in part or in total to increases in screening mammography. The strongest evidence of the incidence in DCIS due to use of screening mammography comes from eight population-based trials of mammography screening. These trials were initiated between 1963 and 1982: the Health Insurance Plan study,134 the Malmo study,149 the Swedish Two-County trial,150 the Edinburgh trial,129 the Stockholm trial,130 the Canadian National Breast Screening Studies 1 and 2,131,132 and the Gothenburg Breast Screening Trial (Table 2).133

Table 2. Population-based screening trials.

Table 2

Population-based screening trials.

The trials consistently reported that less than 20 percent of screen-detected breast cancers were DCIS. The Two-County Study only found a low of 8 percent of breast cancers to be DCIS, while the NBSS-1 found a high 19 percent of breast cances to be DCIS. Thus, all trials found that mammographic screening was more likely to lead to the diagnosis of invasive breast cancer than of DCIS. The Two-County Study observed slightly lower rates of invasive cancer among the screened relative to ususal care (RR 0.95) and significantly higher rates of DCIS among screened relative to usual care RR of screening 1.95 (95 percent CI 1.38; 2.74).51 57 All but the National Breast Cancer Screening trials found mammography to result in significant reductions in breast cancer mortality. An analysis combining the Gothenburg Trial and the Two-County Trial8 defined over-diagnosis as histologically confirmed DCIS detected by active screening that would not have been diagnosed clinically during a woman’s lifetime without screening. This was assessed by comparing the number of cases of DCIS and invasive cancer in the screened population relative to the control. The authors estimated that 15 percent of DCIS cases in the Swedish Two-County trial and 18 percent of DCIS in the Gothenburg Trial represent over-diagnosis and concluded that over diagnosed DCIS did not present a major clinical or public health problem.

The conclusions from the randomized trials are supported by a number of population-based studies from the United States and around the world. Namely, while mammography results in increased detection of DCIS, the number of invasive cancers always outnumbers DCIS cases (Table 3). The impact of screening in these observational studies was assessed using two related definitions: DCIS incidence per 100,000 female population and per 1,000 screned women. Twenty-one U.S. studies reportedthe number of diagnosed case s of DCIS among the number of screened women during a time period of the study (Appendix Table F19).44–46,48–50,52,58,60,66,71,72,85,87–89,91,92,97,99,103 and six studies reported the cumulative incidence of DCIS in the United States per 1,000 screened women (Appendix Table F20).70,72,75,81,88,101 Figure 30 illustrates the relationship of mammography rates, DCIS, and invasive breast cancer in the United States. Invasive breast cancer has not increased significantly since 1987 and has actually declined since 2000. While DCIS increased 200 percent over this period and mammography use increased by almost 250 percent, the increase in mammography use was seen considerably sooner than the increase in DCIS.

Table 3. Diagnosis of DCIS and invasive cancer among screened populations.

Table 3

Diagnosis of DCIS and invasive cancer among screened populations.

Figure 30. Percent change in the age-adjusted incidence of DCIS, invasive breast cancer, and mammography.

Figure 30

Percent change in the age-adjusted incidence of DCIS, invasive breast cancer, and mammography.

The effect of screening programs on incidence of DCIS per 1,000 screening mammograms was studied using data from the Breast Cancer Surveillance Consortium and the National Breast and Cervical Cancer Early Detection Program.72,75,81 Cumulative incidence did not differ among screening programs.72,75,81 The incidence of screen-detected DCIS (0.78 per 10,000 screened, 95 percent CI 0.60; 0.95) was greater than the incidence of nonscreen-detected DCIS (0.13 per 10,000 nonscreened). The same pattern was observed across all age categories (Figure 31). Incidence of DCIS in the United States increased over time as measured with both definitions. The data revealed greater increases over time in incidence per 100,000 population than per 1,000 screened (Figure 32). That is, the incidence of DCIS increased over time, even when the rate of mammography was constant (Figure 33). The rate of screen-detected DCIS was higher in the older age group (1.07, 95 percent CI 0.87; 1.27) compared to women 40–49 years old (0.56, 95 percent CI 0.41; 0.70).72

Figure 31. Cumulative incidence of DCIS per 1,000 mammograms from 1996–1999.

Figure 31

Cumulative incidence of DCIS per 1,000 mammograms from 1996–1999. BCSC - Breast Cancer Surveillance Consortium; NBCCEDP- National Breast and Cervical Cancer Early Detection Program

Figure 32. Time trend in crude annual incidence of DCIS per 1,000 mammograms from January 1997 to December 2003 in women ages 50–69 years (results from Breast Cancer Surveillance Consortium mammography registries).

Figure 32

Time trend in crude annual incidence of DCIS per 1,000 mammograms from January 1997 to December 2003 in women ages 50–69 years (results from Breast Cancer Surveillance Consortium mammography registries). Expon = exponential trend

Figure 33. Annual incidence of DCIS per 1,000 screening mammograms from January 1996 to December 1999 among age catogires of U.S. women depending on screening status (resutls from seven regional mammography registries).

Figure 33

Annual incidence of DCIS per 1,000 screening mammograms from January 1996 to December 1999 among age catogires of U.S. women depending on screening status (resutls from seven regional mammography registries).

There is considerable evidence that the detection of DCIS is greatest at baseline screen. An average annual incidence of DCIS per 1,000 screening mammograms was greater after the first screen for women 50–59 and 70–84 years of age than for subsequent screens (Figure 33).72 Both screening and population-based studies point to increased detection on baseline screen and decreased rates of DCIS detection on followup screens. Though the differences are not large, they do suggest that the greatest increase in incidence will be observed when a population undergoes initial screening and that the increases in incidence based on this initial screen will over estimate population impact for a population undergoing routine screening.

Incidence of different subtypes of DCIS was examined using data from the BreastScreen NSW, an Australian mammographic screening program (Figure 34).76 Incidence of high grade DCIS was greater (4.2 per 100,000, 95 percent CI 3.9; 4.5) than low grade DCIS (1.2 per 100,000, 95 percent CI 1.1; 1.4). Incidence of small tumors less than 2cm was greater (2.1 per 100,000) than for larger DCIS tumors more than 2cm (1.1–1.4 per 100,000).76 Several U.S.-based studies have noted that the incidence of noncomedo DCIS increased substantially while the incidence of comedo DCIS, a high grade, high risk subset, has not increased as dramatically (Figure 35).17,80,82

Figure 34. Cumulative incidence of DCIS by tumor grade and size in Australia (New South Wales Central Cancer Registry, per 100,000 women age standardized to the world population from 1995–2000).

Figure 34

Cumulative incidence of DCIS by tumor grade and size in Australia (New South Wales Central Cancer Registry, per 100,000 women age standardized to the world population from 1995–2000).

Figure 35. Age-adjusted incidence rates of different histological types of DCIS among women ages 30 years, 1980 to 2001 (results from 9 SEER registries in Connecticut, Hawaii, Iowa, New Mexico, and Utah and in the metropolitan areas of Atlanta, Detroit, San Francisco-Oakland, Seattle-Puget Sound).

Figure 35

Age-adjusted incidence rates of different histological types of DCIS among women ages 30 years, 1980 to 2001 (results from 9 SEER registries in Connecticut, Hawaii, Iowa, New Mexico, and Utah and in the metropolitan areas of Atlanta, Detroit, San Francisco-Oakland, (more...)

Several studies examined whether screening had differential impact on DCIS incidence across racial/ethnic groups (Appendix Table F21).70,72,75,81,88,101 Caucasian, Chinese, and Filipino women had the same incidence of DICS (1.6–1.7 per 1,000 mammograms) after adjustment for age, previous mammogram, family history of breast cancer, age at live birth, and BMI.88

Chemoprevention and detection of DCIS in high risk women.

Chemoprevention of DCIS. While several trials have been undertaken that have been used to assess the value of tamoxifen or ralofene for preventing DCIS, the trials, in reality, were designed to assess the value of the agents for preventing breast cancer, with DCIS as a secondary outcome. Several well designed, double blind, RCTs investigated the protective role of tamoxifen on DCIS.105–107 The National Surgical Adjuvant Breast and Bowel Project P-1 study151 examined the protective effect of tamoxifen among high risk women. The study found statistically significant reductions in both DCIS and invasive breast cancer associated with tamoxifen use. The International Breast Cancer Intervention Study enrolled 7,152 high risk women between the ages of 35 and 70 from the United Kingdom, Australia, and New Zealand. The women were randomized to tamoxifen, 20mg/day for 5 years, or placebo.105 The tamoxifen group experienced a 69 percent reduced incidence of DCIS at 50 months (RR 0.31, 95 percent CI 0.12; 0.82) (Figure 36). The protective effect, however, was 4 years after treatment stopped (study month 96) suggesting that the value of tamoxifen for preventing breast cancer or DCIS may not be maintained after treatment ceases.106 The Royal Marsden breast cancer prevention trial107 assigned 2,494 healthy women to oral tamoxifen (20mg/day) or placebo for 8 years. The study did not find a significant protective effect of tamoxifen on DCIS incidence at 13 years of followup. While suggestive, it did not find a statistically significant protective effect for invasive cancer (hazard ratio [HR] 0.78, 0.58–1.04).

Figure 36. Chemoprevention of DCIS with tamoxifen (results from randomized trials).

Figure 36

Chemoprevention of DCIS with tamoxifen (results from randomized trials).

The Study of Tamoxifen and Raloxifene (STAR) trial was a randomized trial of over 19,000 women who were randomized to one of two therapies for preventing breast cancer. Women in the tamoxifen group had half the incidence of in situ breast cancer (lobular carcinoma in situ [LCIS] or DCIS) than women in the raloxifene group (57 versus 81 in situ cancers). However, the study also found with both treatments the risk of invasive breast cancer decreased by half. Offsetting the reduced incidence of DCIS was the observation that the women randomized to raloxifene after 4 years had 36 percent fewer uterine cancers and 29 percent fewer blood clots than the women assigned to the tamoxifen arm.152

The Continuing Outcomes Relevant to Evista (CORE) and Multiple Outcomes of Raloxifene Evaluation (MORE) are randomized double-blind trials examining the impact of raloxefene for preventing invasive breast cancer among post-menopausal women with osteoporosis.153 The CORE trial represents increased followup of the MORE population. The CORE study found significantly reduced incidence of invasive breast cancer associated with raloxifene (HR 0.50) but a nonsignificant increase in the incidence of DCIS among the treated women (HR 1.78). The inconsistent impact of raloxefene on DCIS and invasive breast cancer incidence deserves further investigation and may, ultimately, shed light on the biology of DCIS and invasive breast cancer and factors the control invasive progression.

High risk screening (Appendix Tables 12 and F13). It is well recognized that mammography does not have perfect sensitivity or specificity. As a result, there are ongoing efforts to improve the sensitivity and specificity of screening modalities, particularly for women at high risk of developing breast cancer. One characteristic that is associated with poorer sensitivity of mammography is dense breast tissue. While current guidelines do not recommend screening ultrasound for detection of breast cancer, there is some literature suggesting that ultrasound alone or in combination with mammography might be superior in this case. We found no evidence that ultrasound can improve detection of DCIS in asymptomatic women during population screening programs. The largest U.S. study of 11,130 asymptomatic women who underwent 27,825 screening sessions reported 75.3 percent sensitivity, 96.8 percent specificity, and 20.5 percent positive predictive value of screening ultrasound to detect breast cancer.154 However, the proportion of false-positive results with ultrasound was higher than with mammography.155 Evidence from screening studies in women with radiographically dense breasts suggested that 0.1 percent156 to 0.3 percent157,158 of diagnosed breast cancer cases were diagnosed with ultrasound only. Two studies reported that the specificity of ultrasound is lower in younger women than older women.154,155 In addition to screening mammography, ultrasound can accurately distinguish some solid lesions as benign, reducing the rates of unnecessary biopsy.159,160 The American Cancer Society Guidelines for Breast Cancer Screening found limited clinical evidence for effectiveness or equivalence of ultrasound to screen-film mammography for screening for breast cancer.155

Screening MRI is another option for breast cancer screening. Due to high cost, it is not recommended for routine use but has been explored for women with very high risk, such as carriers of BRCA 1 and 2 genes. Eight prospective case series reported rates of MRI-detected DCIS associated with the BRCA 1 and 2 genes (Appendix Table F22).84,161–167 Cumulative incidence was 1 percent163 or less.84,161,162,164–166

One American study of BRCA1 or BRCA2 mutation carriers of women with less than a 10 percent risk of developing breast carcinoma at 10 years, reported the highest detection rate of DCIS by MRI, 2.4 percent (95 percent CI 0.3; 15.4).167 The studies did not compare detection rate after MRI with other diagnostic procedures. One study compared the predictive value of MRI to mammography to detect breast cancer in women with family history using population based screening in the Memorial Sloan-Kettering breast cancer trials.103 Crude detection rates tended to be higher after mammography (1.2 percent) compared to MRI (0.5 percent). The positive predictive value of MRI was higher (13 percent versus 6 percent) among those with the strongest self-reported family history; the authors concluded that MRI screening should be provided for women with a strong family history of breast cancer.

Finally, the European Group for Breast Cancer Screening consensus statement stated the value of diagnostic ultrasound for targeted examination of both palpable and impalpable breast abnormalities with no evidence to support screening ultrasound in asymptomatic women.168 The American Cancer Society guideline recognized there was insufficient evidence to support the addition to mammography of other screening modalities such as ultrasound or MRI for women at high risk of breast cancer incidence.155

Conclusion. There is ample evidence that the incidence of DCIS is increasing and that the increases are largley due to increased use of screening mammography. Several population-based trials along with other population-based registries also support the conclusion that mammogrpahy is more effective at identifying invasive breast cancer than DCIS. We were unable to find any study that reported both DCIS and inivasive breast cancer that reported detecting more DCIS than invasive breast cancer. Thus, while the increase in DCIS is likely due to screening, the benefits of screening as a means of detecting invasive breast cancer outweigh the increased detection of DCIS.

There is remarkable similarity in risk factors between DCIS and invasive breast cancer with two notable excpetions—first, the age pattern of DCIS and invasive breast cancer are somewhat different. DCIS peaks at a younger age than does invasive cancer. Second, there is no evidence that HRT is associated with increases in DCIS incidnece as it is with invasive breast cancer. Other risk factors including breast density, family history, and history of benign breast disease are similar between invasive cancer and DCIS.

Trials of tamoxifen and raloxefene for breast cancer prevention point to both drugs being effective for preventing invasive breast cancer but tamoxifen being more effective for preventing DCIS. Understanding this effect and how best to prevent all forms of breast cancer deserves further attention.

Question 2. How does the use of MRI or SLNB impact important outcomes in patients diagnosed with DCIS?

Magnetic Resonance Imaging

Post-diagnostic MRI is typically used to guide surgical decisionmaking among the options of breast conserving surgery, mastectomy, and bilateral mastectomies. The differential accuracy of MRI over mammography for accurately identifying these factors defines the value of the technology. Surgical decisionmaking generally takes the following factors into account: multicentric disease, tumor size, and contralateral disease. We analyzed 57 studies181–196 165,166,197–235 that reported the outcomes of breast MRI among patients with established DCIS. Most studies of post-diagnostic breast MRI did not report separate outcomes for invasive breast cancer and DCIS. For our final analysis we excluded those studies. Although this decision limited the number of eligible studies, the patient population of interest was better defined and more generalizable to the specific issue of DCIS. Because these were generally observational studies, many included highly select patients with DCIS who were at greatest risk of having multicentric or extensive disease; these results may not be reflective of all or even most patients with DCIS. We excluded studies when a later publication from the same institution included patients from an earlier study.181,236 237–240 We were unable to find any study that directly compared survival, recurrence, or quality of life for women receiving post diagnostic MRI to no MRI or SLNB versus no SLNB.

MRI for detecting multicentric disease. The presence of multicentric disease is generally considered a contraindication to BCS. Thus, MRI-detected multicentric disease in women with DCIS would be expected to influence treatment recommendations. In a study that included 51 patients with DCIS, Hwang et al. reported that the sensitivity of detecting multicentric disease was significantly higher for MRI as compared to mammography. They estimated MRI to have 94 percent sensitivity compared with mammography that had 38 percent sensitivity (p <0.05).208 Similarly, in a study of 32 patients with DCIS, Menell et al. reported that the sensitivity of detecting multicentric disease was 80 percent for MRI and 40 percent for mammography.199 However, Santamaria et al. studied 86 women with DCIS and did not find the sensitivity of MRI to be significantly better than mammography, although performance of MRI was considerably better than mammography (MRI, 42 percent; mammography, 26 percent; p=.453) (Table 4).223

Table 4. Sensitivity and specificity of breast MRI for detecting multicentric disease.

Table 4

Sensitivity and specificity of breast MRI for detecting multicentric disease.

Menell et al.199 and Hollingsworth et al.229 reported that MRI detected occult multicentric disease at 6.25 percent and 6.3 percent of DCIS patients, respectively. Despite these similarities, variability in the definition of multicentric disease limits comparisons across studies. For example, Hollingsworth defined multicentric disease as a separate focus of cancer more than 5.0cm away from index lesion or discontinuous growth to another breast quadrant,231 while Hwang defined multicentric disease simply as a tumor within at least two quadrants.208

MRI for estimating tumor size. Several studies compared the accuracy of MRI and mammography with histological examination for determining tumor size. The limitations of this comparison group must be acknowledged. Given the growth pattern of DCIS, limitations inherent in tissue processing make histologically-based tumor measurement difficult as 3-diminsional extent of disease is reconstructed using 2-dimensional pathology slides. Thus, pathological examination can overestimate or underestimate tumor sizes, depending on the plane of section. Some authors have argued that MRI measurements may be more accurate than those in the pathology laboratory.231

The results of studies comparing mammography with MRI have not been consistent. In a study of 167 patients with DCIS, Kuhl et al. reported that MRI was not better than mammography in determining size.191 In another study of 24 patients with DCIS, Uematsu et al. reported that MRI was more accurate than mammography in determining extent of DCIS.241 Several studies have evaluated the underestimation and overestimation rates of MRI in determining DCIS size relative to pathological exam (Table 5). Definitions of error were not consistent between studies (+/− 5mm to 10mm), and some studies did not explicitly define what they considered to be an error. For example, in a study of 54 patients with DCIS, Schouten van der Velden et al. reported that MRI overestimated size (defined as >0.5cm) in 38 percent of patients and underestimated size (defined as >0.5cm) in 24 percent of patients.196 In another study of 45 patients with DCIS, Esserman et al. reported that the correlation between MRI and histological size was modest (r=0.55; p=.0001); MRI overestimated size by more than two-fold in 23 percent of patients; MRI underestimated size by half in 9 percent compared to histology.222

Table 5. Overestimation and underestimation of DCIS size by MRI compared with mammography.

Table 5

Overestimation and underestimation of DCIS size by MRI compared with mammography.

MRI for detecting contralateral breast cancer. We found four studies that reported the use of MRI to detect contralateral breast cancer in patients with DCIS (Table 6). In the largest study that included 196 patients, Lehman et al. reported MRI detected occult contralateral breast cancer in five patients (2.6 percent); the sensitivity of detecting contralateral breast cancer was 71 percent.218 Importantly, in this study MRI findings prompted biopsies of the contralateral breast in 18 patients; only five (28 percent) were positive. None of these studies compared the performance of MRI to mammography.

Table 6. Proportion of patients with MRI-detected contralateral breast cancer.

Table 6

Proportion of patients with MRI-detected contralateral breast cancer.

MRI for identifying invasive disease. If MRI could more accurately differentiate between DCIS and invasive cancer, it could alter the surgical treatment of women initially diagnosed with DCIS. We found only one study that evaluated the ability of MRI to identify invasive disease among patients originally diagnosed with DCIS.208 Among 17 patients with DCIS originally diagnosed by core needle biopsy, Hwang et al. reported three patients had invasive breast cancer after definitive surgery; MRI correctly predicted invasive breast cancer in all three patients (sensitivity = 100 percent).208 Hwang estimated the specificity of MRI for detecting invasive breast cancer was 86 percent. After excisional biopsy, the sensitivity of MRI for detecting invasive breast cancer was 75 percent and the specificity was 85 percent. Among all patients, the positive predictive value of MRI for detecting invasive breast cancer was only 43 percent.

Treatment utilization. Nineteen articles reported treatment utilization after diagnostic MRI (Appendix Tables F23 and F24).183,187,191,192,196,199,205,208,210,212,218,221,223,225,227,229–232,234 All articles presented institutional experience performing MRI in DCIS patients (level III evidence). The studies reported descriptive information and did not use strategies to reduce bias. Rather, they reported crude numbers of events in MRI and no MRI groups.

Several studies reported change in treatment decisions based on MRI. Tillman reviewed the medical records of 41 consecutive patients with DCIS who underwent breast MRI from November 1992 through June 2000 prior to planned breast conserving surgery to evaluate the extent to which MRI findings caused any change in the patient’s surgical management.212 The authors reported that MRI simply confirmed information already obtained by mammogram, ultrasound, or clinical examination and did not affect clinical management in 85.4 percent of the patients. Treating surgeons changed local management based on MRI findings in 14.6 percent of the women.212 A study of 32 women treated at Memorial Sloan-Kettering Cancer Center found that MRI findings resulted in changing surgical treatment from breast conserving therapy to mastectomy in 50 percent of women.199 A review of the medical records of 28 women who underwent breast MRI reported that MRI findings changed surgical management for 25 percent of women undergoing pre-surgical MRI.187 In a recent report of 5,596 breast cancer patients (18 percent had DCIS), Katipamula et al. reported that MRI was associated with higher mastectomy rates at the Mayo Clinic.242

Patient outcomes. A single study evaluated whether pre-treatment MRI was associated with rates of local failure among 136 women who underwent BCS followed by radiation therapy at the Hospital of the University of Pennsylvania.183 The rates of local failure were the same (6 percent) among women with or without MRI; the authors concluded that the use of breast MRI was not associated with improvement in outcomes after BCS with radiation.183 The study did not consider changes in treatment strategy as the result of MRI as part of their outcomes evaluation.

Summary. While studies are small, all consistently point to changes in treatment after MRI. These changes are due to differential ability for MRI to detect multicentric and contralateral disease and accurately estimate tumor size.

Sentinel Lymph Node Biopsy

We identified 50 studies that reported experience with SLNB in women with DCIS.98,236–240,243–286 Half of the publications were presented by U.S. academic centers,236,237,243–250,253,256,259–264,267,269,273,275,279,283,285 two studies were conducted in South America,270,271 one in Canada,276 one in Australia,252 and one in Taiwan;257 the rest included women from European countries.

The majority of the studies included middle aged women (median age 50–60 years); few specifically focused on younger (median age <50)237,255,270 or older (median age >60)259 patients.

The authors conducted retrospective review of medical records238,239,252,265,267,270,275,276,284–286 or prospective collection of patient outcomes;98,244,248,249,253,262,268,269,271,282 few reported length of followup240,252,260,264,267,269,273,275,278,279,282 that ranged from 13 months264 to 5 years.252 Only one study reported proportion of loss to followup.264 Sample sizes of the studies (total 7,628 subjects) varied from less than 20 women with DCIS244,258,260,271 to more than 500 patients.240,263,278,283 One article reported the results from a prospective, multi-institutional University of Louisville Breast Cancer Sentinel Lymph Node Study253 that investigated several hypotheses related to SLNB in women with early stages of breast cancer.

The largest series of DCIS women were analyzed in the European Institute of Oncology,240,278 the University of Texas M.D. Anderson Cancer Center,283 and in the database at the H. Lee Moffitt Cancer Center and Research Institute.263 These large academic centers were the basis for more than one publication with different patient outcomes related to SLNB for DCIS; however, we could not exclude the possibility that the same patients were included in more than one of these articles. Two publications compared patient outcomes after SLN and axillary lymph node dissections.256,269

Few studies evaluating SLNB for DCIS include consecutive patients, but rather most report the outcomes of highly selected patients. For example, Yen et al. reported that SLNB was performed on only 35 percent of patients with DCIS.264 Common selection criteria listed by many authors include palpable mass, radiographic mass, large size, mastectomy treatment, high nuclear grade, and suspicion for invasive breast cancer.248,264,274 Patients treated with mastectomy are usually overrepresented in SLNB studies. For example, Meijren et al. reported that 76 percent of patients with DCIS treated with mastectomy underwent SLNB as compared with only 14 percent of patients treated with excision.274 As a result, the published studies are not necessarily reflective of all, or even most, patients with DCIS.

For our final analysis, we excluded several studies for the following reasons:

  1. A later publication from the same institution included patients from an earlier study.236 237–240
  2. SLNB was not performed.252,287,288
  3. The study was a meta-analysis of previously published studies.289
  4. The study did not clearly identify the proportion of patients with DCIS who had SLN metastases.290

We were unable to find any study that directly compared important patient outcomes (survival, recurrence, and quality of life) after SLNB compared with no SLNB.

A review commissioned by AHRQ9 assessed the effectiveness of needle biopsy. The authors synthesized the evidence from 104 studies and concluded that 24 percent of tumors with DCIS identified from stereotactic-guided automatic gun core needle biopsy were found to have invasive breast cancer upon surgical excision (95 percent CI 0.18; 0.32). For stereotactic guided vacuum-assisted core needle biopsy this rate was 13 percent (95 percent CI 0.11; 0.15). Since some patients with an original core needle biopsy of DCIS will have invasive breast cancer identified in the excision or mastectomy specimen, we evaluated the incidence of SLN metastases separately for patients with an original and final diagnosis of DCIS (Tables 7 and 8). The incidence of SLN metastases was greater for patients with an original diagnosis of DCIS (9.8 percent, 95 percent CI 7.6; 12.7)compared with those with a final diagnosis of DCIS (5.0 percent, 95 percent CI 3.6; 6.8)of DCIS. For example, in a study of patients initially diagnosed with DCIS by core needle biopsy, Moran et al. reported that 8.6 percent of patients had SLN metastases.98 However, in this series all patients with SLN metastases had a final diagnosis of invasive breast cancer after excision or mastectomy; thus, no women with a final diagnosis of DCIS had SLN metastases.

Table 7. Incidence of SLN metastases among patients with an original diagnosis of DCIS.

Table 7

Incidence of SLN metastases among patients with an original diagnosis of DCIS.

Table 8. Incidence of SLN metastases among patients with a final diagnosis of DCIS.

Table 8

Incidence of SLN metastases among patients with a final diagnosis of DCIS.

Some studies evaluating the role of SLNB include DCISM, while others include only pure DCIS without microinvasion. Since DCISM may have a higher incidence of SLN metastases, we distinguished DCIS from DCISM in our analysis (Table 9). The incidence of SLN metastases was higher for patients with DCISM (9.3 percent; 95 percent CI 6.0; 14.0) compared with those with DCIS (4.8 percent; 95 percent CI 3.4; 6.7).

Table 9. Incidence of SLN metastases among patients with either DCIS or DCISM.

Table 9

Incidence of SLN metastases among patients with either DCIS or DCISM.

The incidence of SLN metastases and the type of metastases vary according to definitions used. In a multi-institutional study of 470 patients with DCIS, Moore et al. reported that the overall incidence of SLN metastases was 9 percent.275 In this dataset, the incidence of SLN metastases according to AJCC staging was: pN1 (macrometastases), 0.64 percent; pN1 (mic), 0.85 percent; and pN0(i+), 7.70 percent. Using the same dataset but different definitions of SLN metastases yielded slightly different results: routine hematoxylin (H&E), 0.85 percent; serial section using H&E, 4.47 percent; IHC only, 3.83 percent. Whenever possible, we determined the incidence of SLN metastases according to AJCC definitions provided by individual investigators. While many studies267,268,276 defined SLN metastases according to strict AJCC staging, others281 did not use IHC to identify lymph node metastases. Some studies classified SLN metastases as negative, H&E positive, or IHC positive, but did not specify metastasis size.250 In other studies the authors do not distinguish between AJCC stage pN0(i+) and pN1mic.248

The most widely used definition of SLN metastases is the AJCC classification which defines lymph node metastases according to method of detection immunohistochemistry (IHC) and metastasis size. Table 10 lists the incidence of SLN metastas es in studies that defined SLN metastases according to these standards. The incidence of pN1 SLN metastases was 0.9 percent (95 percent CI 0.5; 1.5) in patients with DCIS; 2.3 percent (95 percent CI 0.8; 6.5) in patients with DCISM; and 0.6 (95 percent CI 0.2; 1.6) in the samples that combined DCIS and DCISM. The incidence of pN1(mic) SLN metastases was 1.5 percent (95 percent CI 0.8; 2.8) in patients with DCIS; 3.4 percent (95 percent CI 1.5; 7.7) in patients with DCISM; and 2.6 percent (95 percent CI 0.4; 15.7) in the samples that combined DCIS and DCISM. The incidence of pN0(i+) SLN metastases was 4.2 percent (95 percent CI 2.2; 7.7) in patients with DCIS; 3.5 percent (95 percent CI 1.4; 8.4) in patients with DCISM; and 3.8 percent (95 percent CI 0.7; 18) in the samples that combined DCIS and DCISM. Thus, the incidence of pN1 metastases was very low for patients with pure DCIS.

Table 10. Incidence of SLN metastases according to AJCC staging system.

Table 10

Incidence of SLN metastases according to AJCC staging system.

Since about 15 percent of patients with DCIS identified on core needle biopsy are diagnosed with invasive breast cancer after excision or mastectomy,9 the feasibility and accuracy of SLNB after excision is relevant to decisions regarding surgical management of DCIS. Most studies demonstrate that SLNB is feasible after excision.1,291,292 In a multicenter study of 229 surgeons, Wong et al. reported that the SLN identification rates were similar after core needle biopsy (92.4 percent) and excisional biopsy (92.8 percent).291 However, results from studies evaluating the accuracy of SLNB after excision are not consistent. For example, in the study by Wong et al. the SLNB false negative rates were similar after core needle biopsy (7.9 percent) and excisional biopsy (8.3 percent).291 However, in an analysis from NSABP B-32, Krag et al. reported that the SLNB false negative rate was significantly increased after excisional biopsy compared with core needle biopsy or fine needle aspiration (needle biopsy, 8.1 percent; excisional biopsy, 15.3 percent; p = .0082).1 In this study, the false negative rates were highest for cancers in the lateral portion of the breast, which may make SLNB more difficult.

Although SLNB is minimally invasive and has less morbidity than ALND, the procedure is not risk free. In a prospective Swiss multicenter study, Langer et al. reported the following complications after SLNB alone: lymphedema (3.5 percent), impaired shoulder range of motion (3.5 percent), arm/shoulder pain (8.1 percent), and numbness (10.9 percent).293 In the American College of Surgeons Oncology Group Trial Z0010, Wilke et al. reported that 6.9 percent of patients undergoing SLNB only developed objective evidence of lymphedema.294

Twenty-six studies reported the number of patients who underwent different treatments for DCISafter SLNB (Appendix Table F25).236–240,247,248,252,254,255,257,261,262,264,267,269,273,275–282,285 In some studies axillary lymph node dissection was conducted in all patients with positive SLN,236,239,254,257,277,280,282 while other studies selected patients for further axillary lymph node dissection by the presence of macrometastasis in SLN,276 baseline high risk of metastatic cancer,267,275 or by the discretion of the attending surgeon.262 The studies did not report treatment utilization by positivity of SLN or changes in treatment decisions based on SLNB results. Therefore, the studies describe current practices in the institutions for patients with DCIS who also underwent SLNB rather than examine hypotheses of the association between the results of SLNB and treatment utilization.

Conclusions. The consistent finding that a measurable percentage of women with DCIS on biopsy will be diagnosed with invasive cancer based on full excision suggests that surgical excision of DCIS may be needed to fully evaluate cases for invasive cancer. The findings that some women with confirmed DCIS will have positive SLNB raises questions about whether this seemingly inconsistent finding reflects underdiagnosis of invasive cancer, over diagnosis of positive SLN, or a need to reexamine the presumed association between tumors and nodal involvement. Little data links use of SLNB or positive SLNB with clinical outcomes or treatment changes.

Question 3. How do local control and systemic outcomes vary in DCIS based on tumor and patient characteristics?

We identified 133 publications that addressed the relationship between demographic, tumor or other factors and outcomes of DCIS. The most consistently measured outcomes were local DCIS (72), local invasive cancer (82), local DCIS and invasive cancer (105), contralateral DCIS (20), contralateral invasive cancer (27), combined contralateral DCIS and invasive cancer (44), breast cancer mortality (63), and all-cause mortality (47) (Appendix Table F26). No studies reported chemotherapy use; 16 reported regional recurrence and 44 report distant recurrence. The concept of DCIS recurrence is somewhat challenging, and the literature surrounding this issue is not entirely clear. Technically, a recurrence suggests that the original tumor returned. In contrast, a new primary invasive cancer or new DCIS refers to a new tumor arising in the same or a different area of the ipsilateral (same side) or contralateral breast. Few studies differentiate between recurrence and new primary invasive cancer or DCIS. Rather, in most cases, these are combined and variously called ‘recurrence’ or ‘local DCIS.’ Rarely, if ever, are ipsilateral tumors carefully examined to differentiate between these two etiologies. Even clinically, this is rarely fully explored and not clearly helpful with decisionmaking. For the purposes of this report, we will follow the language of the literature and consider ‘recurrence’ to mean DCIS or invasive cancer in the same breast as the original tumor unless otherwise specified.

At 10 years following DCIS diagnosis, overall breast cancer mortality consistently is less than 2 percent.295–297 In official publications, the SEER registries report 0 percent breast cancer mortality after 5 years, reflecting the belief that there is no mortality from DCIS unless there is an invasive recurrence or new invasive primary tumor, in which case the mortality would be attributed to the recurrence or new tumor.4 Ernster5 estimates 0.7 percent breast cancer mortality within 5 years and 1.9 percent within 10 years for women diagnosed between 1984 and 1989. Ernster also reports that breast cancer mortality declined significantly between 1978–1983 and 1984–1989 (10 year mortality at 10 years 3.4 percent versus 1.9 percent).

Recurrence of both DCIS and invasive disease is the most common ongoing consequence for women diagnosed with DCIS. Estimates of 5 or 10-year recurrence rates are remarkably unstable across studies ranging from 2.4–15 percent for 5-years to 10–24 percent for 10-year recurrence. Estimates from cancer registries such as SEER are somewhat problematic since registries, by design, do not collect information on recurrence but do collect information on new primaries. While an invasive cancer after DCIS should be reported to the registry, some confusion likely remains. When both 5- and 10-year outcomes are reported for the same cohort, it is interesting to note that in some cases, such as Vicini, there is relatively little increased risk in years 5–10 beyond what was experienced in the first 5 years.298 For example, Vicini reports a small case series where the 5-year rate of local DCIS or invasive recurrence is 10.2 percent and at 10 years the rate is 12.4 percent.298 In other cases, however, there is a large difference in risk between 5 and 10 years. This raises questions about whether risk of recurrence is stable over time, whether it increases or decreases.

Contralateral DCIS disease is a less common occurrence with an incidence estimated to be up to 1.7 percent after 7 years followup. When combined with invasive contralateral breast cancer, incidence rises to up to 8 percent after 10 years. Of note, the five studies299–303 that report both contralateral DCIS and contralateral combined invasive cancer and DCIS point to between one-third and three-quarters of the incidence attributed to contralateral invasive tumors. Gao304 reports a steady increase in the cumulative incidence contralateral breast cancer in the 20 years following DCIS diagnosis. Over time, however, the 5-year incidence, declines slightly (Figure 37).

Figure 37. Contralateral breast cancer with time since DCIS diagnosis.

Figure 37

Contralateral breast cancer with time since DCIS diagnosis.

Local recurrence is the most adverse outcome experienced by women receiving treatment for DCIS. While somewhat beyond the scope of this report, several small studies provide some evidence of survival after local recurrence. Solin reports on the experience of 42 cases with local recurrence and estimated an actuarial 5-year breast cancer mortality rate of about 16 percent.305 Similarly, in a multi-institutional cohort, 15 women who received treatment for DCIS experienced a local recurrence and received salvage treatment. After a median of 4.4 years 14 of these women were alive.306 Thus, while survivable, local recurrence is serious and preventing local recurrence is clearly preferable.

Tumor Characteristics

Positive surgical margins. Positive surgical margins are consistently associated with increased DCIS and invasive breast cancer recurrence (Figure 38).297,298,307–322 Likewise, two reports from RCTs pooling across treatments found a similar effect.323,324 There was, however, considerable variability across studies in terms of how margins were defined or classified. For example, some studies classified margins as ‘free’ or ‘involved’325,326 while others use more precise measures such as <1mm.327,328 We excluded one study329 because we could not reproduce their significance estimates or conclusions.

Figure 38. Crude odds of local DCIS or invasive carcinoma by margin status in women with DCIS.

Figure 38

Crude odds of local DCIS or invasive carcinoma by margin status in women with DCIS.

Subgroup analyses from two RCTs both reported increased risk of local recurrence in women with positive margins after breast conserving surgery.295,323 For example, the National Surgical Adjuvant Breast and Bowel Project324 reported that women with positive margins after breast conserving surgery had higher risk of local DCIS or invasive cancer than women without positive margins (84 percent increase).324 After a median of 10.5 years of followup, the study reported that women with involved surgical margins had higher risk of ipsilateral recurrence after adjustment for treatment and all other predictors of recurrence (HR 2.06 <.001).6

We synthesized the evidence separately from observational studies of better quality that reported multivariate adjusted estimates of the association between patient outcomes and margin status (14 studies) (Table 11).297,298,308–310,312,313,315,316,318–321,330 The majority of such studies reported a positive significant association between positive margins and recurrence. Other studies reported a nonsignificant increase in the odds of local recurrence in women with involved margins after lumpectomy with or without adjuvant radio or chemotherapy 316 and increased risk of local recurrence in women with close or involved margins after lumpectomy or mastectomy.315

Table 11. Adjusted relative effect of margin on patient outcomes.

Table 11

Adjusted relative effect of margin on patient outcomes.

An analysis of adjusted relative risk (Figure 39)297,320,321 suggests risk of local recurrence is reduced with larger widths of negative margins. Margins of 10mm or more were associated with the largest reduction (98 percent) in the risk of local recurrence, while no differences were seen using a cut off of 2 or 4mm.

Figure 39. Impact of negative margin width on local DCIS or invasive recurrence–multivariate adjusted estimates, pooled with random effects.

Figure 39

Impact of negative margin width on local DCIS or invasive recurrence–multivariate adjusted estimates, pooled with random effects.

Tumor size. The association between tumor size and patient outcomes was examined in two RCTs295,331 and 39 observational studies296,297,301,309–312,314–318,320,327–330,332–352 (Table 12). In general, larger tumors were associated with higher rates of local DCIS and invasive recurrence than smaller tumors,296,311,312,316–318,320,337,338,343,347 though many of the estimates were not statistically significant.295,296,316,327–329,331,333,337,338,347 Estimates generally classified tumors less than 20mm as ‘small’ though some320 defined small as <5mm. A study of 89 women failed to find tumor size to be associated with an increased risk in breast cancer mortality; however, the HR of 2.90 pointed to importantly increased risk.338 There was no consistent finding of an association between tumor size and contralateral DCIS,337 contralateral DCIS or invasive carcinoma,337,345 or contralateral invasive carcinoma.337,338,347 A single study examined the association between tumor size and distant metastases and failed to find a significant association.334 One study found that the odds of all events350 were significantly greater for women with large versus small tumors (OR 11.388, 95 percent CI 1.752; 74). One case series of 455 nonrandomized patients treated with excision alone320 reported a significant increase in relative risk of local recurrence by 21 percent per 1mm increase in tumor size (RR 1.21, 95 percent CI 1.1; 1.34).320

Table 12. Association between tumor size and patient outcomes.

Table 12

Association between tumor size and patient outcomes.

Grade. The association between tumor grade and patient outcomes was reported in 39 studies (Table 13).295,296,306,307,309–313,315–317,320,321,323,325,327,329,330,335,339–343,345,347–349,351,353–361 While labeled somewhat inconsistently, tumors assigned a higher pathological or nuclear grade (3) have consistently higher probably of local DCIS or invasive recurrence than those at intermediate or low grade (2 or 1). Two studies, each with less than 300 women, examined the association between tumor grade and mortality. The European Organization for Research and Treatment of Cancer Trial 10853 demonstrated that women with high grade DCIS treated with lumpectomy plus radiation had a 716 percent increase in relative risk of all cause mortality compared to women with low grade DCIS (RR 8.16, 95 percent CI 1.02; 65.252).357 The association was of similar magnitude but not statistically significant for women treated with lumpectomy alone. The study did not observe increased risk of mortality for intermediate grade DCIS compared to low grade.357 A multi-institution observational study from the United States and Europe of 172 women treated with lumpectomy plus radiation failed to find a significant association between crude odds of death and tumor grade.325 The apparent lack of association between tumor grade and breast cancer mortality could be due to a lack of effect or low power given the overall, low mortality associated with DCIS.325,341 Two studies—one RCT and one observational study—failed to find a consistent association between DCIS grade and distant metastases.325,357 No study found an increased risk of contralateral cancer associated with tumor grade.345,347,356 A single study using SEER cancer registry data found a slight but not statistically significant increase in local or contralateral invasive cancer (HR 1.2) associated with high versus low tumor grade.347 Three of three observational studies reporting any recurrence found that women with high grade DCIS had increased rates of any recurrence relative to women with low grade DCIS.348,351,358 The study that reported multivariate adjusted analysis demonstrated a 122 percent increase in risk of any recurrence in women with high versus low grade DCIS (2.22, 95 percent CI 1.02; 4.76).358 The rates of local invasive recurrence tended to be higher in women with high grade DCIS in all six observational studies that examined this association.296,316,329,347,354,356

Table 13. Association between tumor grade and patient outcomes.

Table 13

Association between tumor grade and patient outcomes.

Comparisons of intermediate (2) versus low (1) grade were much less consistent. While several studies failed to find statistically significant associations between intermediate and low grade tumors,296,310,312,347 Kerlikowske322 found significant increased risk of recurrence for grade 2 versus grade 1 tumors in a cohort of 1,036 women treated with lumpectomy alone.

Millis362 noted that 84 percent of recurrent lesions were of the same grade as the primary DCIS. For recurrent DCIS they observe a kappa of 0.679, while with invasive recurrences the kappa was lower at 0.241; however, almost all of the invasive and DCIS recurrences were associated with high grade lesions (76 percent and 75 percent, respectively). Overall, the studies suggest that the difference between grades 2 and 1 may be less important than the difference between grade 3 and grades 2 and 1. However, Barnes363 noted that the percentage of low grade tumors (i.e., grade 1) was stable between 1979–2000 and 2001–2002, while the percentage of intermediate grade declined (28.1 percent versus 22.7 percent) and high grade tumors increased (62.5 percent versus 68.1 percent). This may point to moderate stage shift. Of note, Li found no association between pathologic grade and contralateral invasive cancers.347

Architecture. The most commonly measured architectural feature of DCIS is comedo necrosis. Noncomedo DCIS includes cribriform, micropapillary, and solid types. Comedo necrosis is consistently and strongly associated with increased risk of local DCIS or invasive cancer with hazard ratios generally above 2.0. and as high as 9.3 (Table 14).296,311,312,315,320,324,337,343,347,364 For example a large analysis of the SEER database 347 demonstrated a 30 percent increase in relative risk of local invasive recurrence (adjusted HR 1.4, 95 percent CI 1.1; 1.7) in women with comedo versus noncomedo DCIS. Warren316 and Sahoo311 both reported no increased risk of local DCIS or invasive cancer recurrence associated with comedo necrosis (RR 0.9 and 0.7, respectively). Li found women with comedo necrosis were at slightly reduced risk of contralateral invasive recurrence.347 No study reported a significant association between comedo and noncomedo DCIS and all cause mortality,325,365 breast cancer mortality,325,366 contralateral invasive carcinoma,347 or all events.334 Only one study325 of three studies325,334,366 found a significant increase in odds of metastasis in women with comedo necrosis (OR 8.609, 95 percent CI 1.038; 71.387).325

Table 14. Association between architecture and patient outcomes.

Table 14

Association between architecture and patient outcomes.

Comparisons between other architectural groups are rarely reported and are somewhat inconsistent. For example, Fisher295 reported increased risk of DCIS or invasive recurrence for women with solid tumors compared with cribriform (RR 2.41), while Bijker323 reported increased risk of cribriform versus clinging/microcapillary tumors (RR 2.39) and for solid/comedo versus clinging/microcapillary tumors (RR 2.25) but didn't compare solid with cribriform to allow for comparisons between the two studies. Smith296 reported a slight, nonsignificant increased risk of local DCIS or invasive recurrence associated micropapillary versus not (HR 1.41) and a strong but not statistically significant decreased risk associated with cribriform versus not (HR 0.27).

Women with solid or cribriform tumor when compared to micropapillary had the same rates of contralateral DCIS, any contralateral cancer,337 or contralateral invasive carcinoma.337,347 Odds of any recurrence did not differ in women with solid versus micropapillary DCIS 301or cribriform versus micropapillary,301 DCIS and by 30 percent (adjusted HR 1.3, 95 percent CI 1; 1.7) in women with papillary versus not specified DCIS.347 A large SEER-based study reported a significant increase by 100 percent (adjusted HR 2, 95 percent CI 1.01; 3.99) in risk of local DCIS recurrence in women with papillary versus not specified DCIS.296 RCTs demonstrated a significant increase in relative risk of local DCIS or invasive recurrence by 139 percent (RR 2.39, 95 percent CI 1.41; 4.03) for cribriform versus micropapillary DCIS and of 125 percent (RR 2.25, 95 percent CI 1.21; 4.18) in women with solid or comedo versus micropapillary DCIS,323 or by 141 percent (RR 2.41, 95 percent CI 1.28; 4.52) in women with solid versus cribriform DCIS.295

Microinvasion. DCIS with microinvasion represents a few isolated tumor cells or clusters of cells infiltrating the periductal stroma. The clinical significance of DCISM is somewhat controversial. Some of these cases are noted as DCISM, some are considered to be DCIS, others invasive cancer. Many publications explicitly note the presence of DCISM while others do not comment on DCISM. The association between microinvasion and patient outcomes was inconsistent in the direction and magnitude across the single randomized trial357 and three of four observational studies342,345,367,368 that compared cases of DCIS with and without microinvasion (Table 15). While not all are statistically significant, all but one reported increases in adjusted risk of local DCIS or invasive carcinoma in women with microinvasion relative to without. The statistically significant study reported a HR of 8.1 associated with microinvasion (95 percent CI 1.2; 53).367

Table 15. Association between microinvasion and patient outcomes.

Table 15

Association between microinvasion and patient outcomes.

Necrosis. One observational study examined the association between mortality or distant metastases and the presence of necrosis and did not find a significant association (Table 16).325 Two observational studies examined the association between contralateral cancer and the presence of necrosis and did not find a significant association.337,345 Three observational studies showed a positive tendency between necrosis and worse rates of any recurrence301,348,358 but only one found a significant association.301 Three observational studies329,337,364 showed that women with necrosis had increased rates of local DCIS recurrence, but only two reported a significant increase by 63 percent364 or 258 percent.337 The association was more evident for local invasive carcinoma; the largest study of 23,547 women with DCIS from the California Cancer Registry showed a 93 percent increase in local invasive cancer in women with necrosis (IRR1.93, 95 percent CI 1.28, 2.91).364 The association between necrosis and local DCIS or invasive cancer recurrence differed depending on the treatments women had. The association was not significant after mastectomy369 or skin-sparing mastectomy,348 inconsistent in direction and significance after lumpectomy plus radiation,306,311,360,369,370 and in studies that combined all treatment together in analysis.312,315,316,329,335,339,345 Women after lumpectomy had an increased risk of local DCIS or invasive recurrence by 115.8 percent (pooled RR 2.158, 95 percent CI 1.263 3.687, I2 25 percent).320,337,343,369

Table 16. Association between necrosis and patient outcomes.

Table 16

Association between necrosis and patient outcomes.

Van Nuys Index. The Van Nuys Index is scored from 4–12 based on four different predictors of local breast recurrence: tumor size, width of negative margin, pathologic classification, and patient age.371 Each predictor is scored from 1–3. The index measures post-surgical risk of events (since surgical margins comprise one-quarter of the score).

The association between patient outcomes and Van Nuys risk category was examined in 15 observational studies (Table 17).317,336,341,343,349,350,352,358,371–377 Comparison of studies reporting Van Nuys Index is complicated because numerical scores are not consistently categorized across studies. Some studies applied the exact Van Nuys criteria;317,336,339,343,349,350,352,372,373,375,377,378 others used the summary index (USC/Van Nuys Prognostic Index) adding age.349,350,371,377 Some studies included age, grade, and tumor size but not surgical margins,376 calculated tumor size from mammographic lesion,358 or modified cut offs for nuclear grade (low=1, intermediate=2, high=3) and margin (>1mm score=2, ≤1mm score=3).374

Table 17. Association between predicted Van Nuys Index risk categories and patient outcomes (results from observational studies).

Table 17

Association between predicted Van Nuys Index risk categories and patient outcomes (results from observational studies).

Women at the highest risk category of Van Nuys index (10–12) had 224 percent greater odds of mortality than women in the 4 to 6 risk category.350 Breast cancer mortality was examined in four studies;350,371–373 one found a significant positive association with greater predicted risk (OR 8.61, 95 percent CI 1.06; 70.17) in women with a Van Nuys score of 10 to 12 compared to those scores of 4 to 6.350 Similarly, Asjoe found that the odds of any recurrence were significantly greater in women with a Van Nuys score of 10 to 12 relative to 4–6 (OR 7.58, 95 percent CI 2.17; 26.55) but not for women with a Van Nuys index score of 7–9 relative to 4–6.349

Multi-focal disease. While rarely precisely defined, two studies reported multifocal disease associated with increased risk of DCIS and invasive cancer recurrence.295,321 Similarly, a small case series (121 women) reported a diffuse growth pattern to be associated with a nonsignificant increased risk of DCIS or invasive recurrence.361,379

Estrogen and progesterone receptor status. Nine studies investigated the association between ER status and patients outcome (Table 18).312,313,330,342,351,379–381 SEER-registry-based analysis shows that less than 14 percent of DCIS cases have ER status tested.80 Thus, studies of ER status and DCIS outcomes are generally limited to small studies, often including approximately 100 cases. Generally, all are consistent in their findings that positive estrogen receptor status is associated with reduced likelihood of local DCIS or invasive recurrence, although few of the associations are statistically significant.379,380,382 For example, the Population-based Regional Tumor Registry in Lund, Sweden, reported their experience with 187 patients found decreased risk of recurrence for women whose tumors were ER positive or unknown compared to ER negative (HR 0.71 and 0.68, respectively).379 Few studies report the association between estrogen receptor status and mortality. Bijker examined the concordance between primary DCIS and recurrence and found a kappa of 0.9 for estrogen receptor status.383 It is notable that the NSABP-35, a trial of whether aromitase inhibitors prevent recurrent DCIS or invasive cancers, is limited to women with ER positive tumors. This trial may be a signal that ER testing for DCIS might become more widespread.384

Table 18. Association between ER status and outcomes.

Table 18

Association between ER status and outcomes.

Barnes363 evaluated 119 consecutive tumors and noted that there is a strong association between the presence of comedo necrosis and estrogen receptor negativity with 73 percent of all tumors being ER+ but only 57 percent of comedo tumors were ER+. A similar negative association was observed between ER positivity and higher tumor grade. The study found that only 64 percent of high grade tumors were ER positive.

Seven studies investigating the association between PR status and patient outcomes showed a tendency toward less local DCIS or invasive cancer recurrence in PR-positive women (Table 19).330,342,351,379–381 One study reported p-value only and is not summarized here.385 However, only one nested case control study within a population-based cohort in Australia reported a significant reduction by 60 percent (adjusted OR 0.4, 95 percent CI 0.2; 0.9)381 in odds of local recurrence in PR positive patients. In contrast, the association between PR status and any recurrence was opposite in direction and neither study achieved statistical significance.351,380

Table 19. Association between progesterone receptor (PR) status and outcomes.

Table 19

Association between progesterone receptor (PR) status and outcomes.

Her2Neu. The relationship between Her2 (human epidermal growth factor receptor-2) positivity and recurrence was only studied in relatively small DCIS studies of 129 patients or less (Table 20).380,386 Consistently, investigators have found women with Her2 positive DCIS were at higher risk of recurrence. Barnes reported that 65 percent of tumors were positive for Her2 expression. They concluded that coexpression of Her2 and Her4 was associated with reduced recurrence compared with Her2 only tumors. The importance of Her2 positivity is highlighted by a study by Bijker which found a kappa of .75 between Her2 positivity on initial DCIS and recurrence.383 Her3 and Her4 have only been evaluated in a single study.

Table 20. Association between HER status and local DCIS or invasive carcinoma.

Table 20

Association between HER status and local DCIS or invasive carcinoma.

Calcification. In multiple reports from the same institution using a moderate sized cohort, (132–148 subjects),298,318,370,387 the lack of calcification was strongly associated with DCIS or invasive carcinoma recurrence (HR 3.57–4.55 calcification versus no calcification). The studies did not classify calcifications based on their form, such as fine/granule, etc.

Characteristics of Women

Age. Younger age at diagnosis is a consistent adverse prognostic factor for DCIS outcomes.

Women over age 40 or 50 consistently have a lower risk of DCIS or invasive recurrence than younger women,297,309,310,312,314–316,322–324,347,364,388 with many studies reporting relative risk around 0.5 and one study reporting the relative risk to be as low as 0.12.310 It is less clear whether the age-related disadvantage is attenuated when comparing middle aged and older women. For example, Innos reported similar recurrence rates between women between 50 and 65 and those over 65.364 Likewise, Li found recurrence rates for women between 50–59 and 60–69 or 70+ to be equivalent.347 Vargas,307 Vicini,298,318,370 and Smith296 modeled age as a continuous variable and found the relative risk of local DCIS or invasive recurrence to decline by approximately 0.95 for each year of age.

Innos reported contralateral DCIS to be highest in women <40 compared to women 50–65 and did not find significant increased risk of contralateral DCIS for other age groups.364 In contrast, Li found increased risk of contralateral invasive cancer to be higher in older women.347

All-cause mortality, however, is consistently lower in younger women than older women.80,389

Consistent with the increased risk of recurrence in younger women, three studies found pre-menopausal women to face higher risk of recurrence than post-menopausal women.309,322,333

Race. Surprisingly few studies report racial differences in DCIS outcomes. SEER-based studies report higher all-cause mortality among African American women than white women diagnosed with DCIS,389 higher breast cancer mortality for African American women than white women,80 and higher nonbreast cancer mortality for African American women than white women80 The analysis by Deshpande et al.390 showed that the mortality disadvantage for African American women was maintained at all age groups. DCIS recurrence among different racial subgroups was compared in five articles that analyzed SEER data296,316,347,376,389 and several others.322,364 Three of the SEER analyses adjusted for clinical prognostic variables, including tumor size, grade, or necrosis296,316,376 and found no differences in local DCIS or invasive carcinoma recurrence, local DCIS recurrence, or local invasive carcinoma recurrence in race subgroups. Two SEER-based papers adjusted for age, year, tumor registry, and treatments but not tumor characteristics.347,389 Those papers reported worse outcomes among African American women compared to whites with DCIS. The papers found overall mortality to be 35 percent higher (RR 1.35, 95 percent CI 1.12; 1.62) in African American versus white women with DCIS.389 African American women had higher rates of local invasive carcinoma recurrence (RR 1.5 95 percent CI 1.2; 2), contralateral invasive carcinoma (RR 1.3, 95 percent CI 1; 1.7),347 or any invasive carcinoma (RR 1.4, 95 percent CI 1.2; 1.7).347 Risk of advanced invasive carcinoma, stage III/IV was 170 percent in African American versus white women (RR 2.7, 95 percent CI 1.7; 4.4).347 These findings point to differences in tumor characteristics such as size, grade, and necrosis as important explanatory factors for the observed poorer outcomes among African American versus white women. The findings also underscore the importance of tumor characteristics that remain after controlling for treatment.

Patient outcomes for Asians or Asian-Pacific Islanders were compared to whites in five articles.322,347,364,376,389 The analysis that adjusted for age and treatment did not find difference in any outcomes: three studies in local invasive cancer recurrence,364 one study in contralateral invasive cancer, one study in any DCIS or invasive cancer recurrence,364 any invasive cancer, and mortality. Asian women diagnosed with DCIS had lower mortality rates than white women.389

Patient outcomes in white Hispanics were compared to whites in four articles.322,347,364,376,389 The analyses adjusted for age, treatment, and, in some cases, histology did not find difference in local DCIS recurrence,364 local invasive cancer recurrence,296,347,364 contralateral invasive cancer, any DCIS or invasive cancer, any invasive cancer, all, stage I, or stage II. However, risk of advanced invasive cancer, stage III/IV was 130 percent higher in Hispanic versus white women with DCIS (RR 2.3, 95 percent CI 1.1; 4.8).347 The studies did not report mortality.

Patient outcomes comparing American Indians to whites were reported in only one article.389 The study includes only 82 American Indian DCIS cases and did not find statistically significant differences in mortality. The small number of cases included in the analysis limits the interpretability of these Native American comparisons.

Mammographic density. Two studies examined outcomes of DCIS associated with mammographic density.391,392 They did not classify mammographic density in the same way, which somewhat limits comparability. Habel, who classified density as a percent, only found an association between mammographic density and local DCIS or invasive recurrence when comparing women with ≥75 percent to <25 percent.391,392 Habel, also reported high mammographic density associated with contralateral disease recurrence (RR 3.4).391

Reproductive history. Few studies examine the association between reproductive history and DCIS outcomes. Habel found no association between younger age at first birth, parity, or hormone replacement therapy and DCIS or invasive cancer recurrence but did find a slight benefit to older age at menarche.333 Oral contraceptive use was reported in two studies.322,333 Neither reported a statistically significant outcome; one reported a history of oral contraceptive use to be a favorable prognostic factor, the other associated with slight increased risk (1.4).

A single cohort of 709 women from western Washington333 is the sole source of information on the prognostic value of several DCIS risk factors. While small, the study does report expected associations between tumor size, comedo necrosis, and BMI. The study reported a nonsignificant association between some (versus no) weekly alcohol consumption and reduced risk of recurrence. Likewise, they found a nonsignificant trend toward decreased risk of DCIS or invasive cancer recurrence and use of oral contraceptives and a nonsignificant increased risk of DCIS or invasive cancer recurrence associated with hormone replacement therapy that did not depend on duration of hormone replacement therapy use or formulation. This study found no consistent association between age at first birth and DCIS or invasive carcinoma recurrence.

Family history. The association between positive family history and DCIS or invasive breast cancer recurrence was reported in four studies.309,314,322,333 All found a positive family history to be associated with increased risk, though not all effects were statistically significant.

Comorbidity. Two studies reported the association between comorbidity and DCIS outcomes. Warren found women with one or more comorbidities were more likely to experience a local DCIS or invasive cancer recurrence than women with no comorbidities (RR 1.62).316 Smith,296 however, found no increased risk of DCIS or invasive cancer recurrence when comparing women with no comorbidities to one or to two to nine comorbidities.

Year of diagnosis. The association between patient outcomes and the year of DCIS diagnosis was examined in four observational studies.5,297,344,364 Women diagnosed with DCIS after screening mammography became common (1984–1989, 5,547 women in SEER database) compared to those diagnosed in 1978–1983 (1,525 women in SEER database) had a 40 percent reduction in adjusted relative risk of breast cancer death.5 The 10-year breast cancer standardized mortality rate in women with DCIS declined from 3.4 (95 percent CI 2.4; 4.5) before screening mammography was common to 1.9 (95 percent CI 1.5; 2.3) after wide implementation of breast cancer screening.5 A large California Cancer Registry-based study evaluated whether the standardized incidence ratio for a primary breast cancer among women with DCIS compared to the general population changed between 1988–1993 and 1994–1999. The study reported the standardized incidence ratio was unchanged (1.4 versus 1.3) in two time intervals.364 A European study of 1,640 DCIS cases analyzed the rates of local recurrence before and after implementation of the clinical guidelines for management of breast cancer.344 The rates of local DCIS or invasive recurrence reduced from 9.6 percent in 1992–1995 to 2.9 percent in 2000–2003. However, there was no significant association between adherence to the guidelines and local recurrence.344 Finally, a multisite study found the rates of local failure were unchanged over time.297 In summary, while observational studies suggested reduction in breast cancer mortality after implementation of mammographic screening in the United States, the rates of local recurrence and contralateral breast cancer remain unchanged over this same period.

Summary

In general, few of the risk factors for DCIS or breast cancer incidence are also associated with outcomes following DCIS diagnosis. However, the majority of important prognostic factors for DCIS outcomes are also prognostic factors for invasive breast cancer outcomes (Table 21). Beyond factors that are routinely measured by cancer registries, many of the factors reviewed in this report rely on the findings of a single cohort of 709 women from western Washington333 as the sole source of information on the prognostic value of several DCIS risk factors. While small, the study does report expected associations between tumor size, comedo necrosis, and BMI. The recurrence rates, however, are higher (31 percent) than reported by many studies (e.g., 10 percent). Thus, there is a need for larger population-based studies of the relationship between tumor markers and patient characteristics on outcomes after DCIS diagnosis.

Table 21. Comparison of major prognostic factors between DCIS and early stage invasive breast cancer.

Table 21

Comparison of major prognostic factors between DCIS and early stage invasive breast cancer.

Question 4. In patients with DCIS, what is the impact of surgery, radiation, and systemic treatment on outcomes?

We identified five randomized trials that addressed the value of radiation therapy (Table 22) or tamoxifen for treatment of DCIS. Of note, we were unable to find any randomized trials comparing BCS plus radiation therapy with mastectomy analogous to the NSABP-B06 trial for invasive breast cancer. In addition to information from randomized trials, we identified 133 publications of 64 observational studies o (i.e., nonrandomized studies) that address the impact of treatment on DCIS outcomes (Appendix Tables F26-F33). The most consistently measured outcomes were ipsilateral DCIS, ipsilateral invasive cancer, combined ipsilateral DCIS and invasive cancer, contralateral DCIS, contralateral invasive cancer, combined contralateral DCIS and invasive cancer, breast cancer mortality, all-cause mortality, chemotherapy use, local recurrence, regional recurrence, distant recurrence, and other outcomes (Appendix Table F26).

Table 22. Summary of characteristics of included RCTs.

Table 22

Summary of characteristics of included RCTs.

For the purposes of this report, we consider BCS, lumpectomy, and wide local excision to be analogous terms.

Breast Conserving Surgery With Radiation Versus Without

In randomized trials including NSABP-17 and the European Organization for Research and Treatment of Cancer (EORTC) randomized phase III trial 10853, whole breast radiation therapy following BCS is associated with a reduction of local DCIS or invasive carcinoma recurrence but no impact on breast cancer mortality or total mortality (Table 23). The studies consistently found whole breast radiation therapy to be associated with a reduced incidence of local DCIS recurrence and local invasive carcinoma. While statistically significant, the number of events prevented per 1,000 treated women is typically less than 10 percent (Table 24).

Table 23. Association between treatment options for DCIS and patient outcomes from RCTs by trial.

Table 23

Association between treatment options for DCIS and patient outcomes from RCTs by trial.

Table 24. Events reduced by treating 1,000 people with radiation after breast conserving therapy (statistically significant effects only).

Table 24

Events reduced by treating 1,000 people with radiation after breast conserving therapy (statistically significant effects only).

Two studies323,324 found that while radiation therapy had a similar effect on recurrence between those with positive and negative margins, the adverse prognostic effect of positive margins remained after RT (HR 1.84;357 RR 1.84324).

Likewise, while Holmberg331 and Fisher295 reported similar effectiveness of RT regardless of tumor size, RT did not completely eliminate the increased risk associated with larger versus smaller tumors (Appendix Table F34).

Multiple observational studies report lower rates of local DCIS or invasive cancer for women undergoing BCS+RT over BCS alone,296,307,308,314,316,319,321,333,338,347,358,396 though not all report statistically significant patterns. Observational data from Sweden338 show a lack of mortality benefit associated with BCS+RT compared to BCS alone, while a single study389 did find women receiving RT had lower all-cause mortality.

While generally low level, there is no evidence that breast conserving surgery plus radiation is more or less effective than breast conserving surgery without radiation in the presence or absence of adverse prognostic factors. This lack of differential effect can be seen for the most important prognostic factors, including grade, tumor size, involved margins, and comedo necrosis. (Table 2526).

Table 25. Impact of tumor grade on the effectiveness of lumpectomy plus radiation vs. lumpectomy alone.

Table 25

Impact of tumor grade on the effectiveness of lumpectomy plus radiation vs. lumpectomy alone.

Table 26. Impact of necrosis on the effectiveness of lumpectomy plus radiation vs. lumpectomy alone.

Table 26

Impact of necrosis on the effectiveness of lumpectomy plus radiation vs. lumpectomy alone.

Mastectomy

While not studied in a randomized fashion, several observational studies (Appendix Tables F35F37) compared outcomes between mastectomy and BCS or BCS+RT. They found women undergoing mastectomy (Appendix Tables F38F39) were less likely than women undergoing lumpectomy (Appendix Table F40) or lumpectomy plus radiation (Appendix Table F41) to experience local DCIS or invasive recurrence.310,315 Women undergoing BCS alone were also more likely to experience a local recurrence (Appendix Tables F42F44).310,315,338 We found no study showing a mortality reduction associated with mastectomy over BCS with or without radiation. It is possible, however, that low statistical power is an important factor behind this apparent lack of benefit. Since the breast cancer mortality after DCIS diagnosis is so low, it is possible that few studies have included sufficient numbers of cases to support identification of a mortality benefit. Selection bias may also contribute to the apparent lack of benefit for mastectomy in observational studies. Clinically larger, multicentric, and more problematic tumors will be more likely to be treated with mastectomy than BCS. These tumors are also more likely to recur and are more often associated with breast cancer mortality. Thus, equal mortality in spite of differences in severity may be masking a clinically superior treatment.

While generally low level, there is no evidence that mastectomy is more or less effective than BCS plus radiation in the presence or absence of adverse prognostic factors. This lack of differential effect can be seen for the most important prognostic factors, including grade, tumor size, involved margins, and comedo necrosis (Tables 2731).

Table 27. Influence of architecture on mastectomy effectiveness.

Table 27

Influence of architecture on mastectomy effectiveness.

Table 28. Impact of grade on the effectiveness of mastectomy vs. lumpectomy.

Table 28

Impact of grade on the effectiveness of mastectomy vs. lumpectomy.

Table 29. Association between treatment and patient outcomes, stratified by architecture.

Table 29

Association between treatment and patient outcomes, stratified by architecture.

Table 30. Impact of necrosis on the effectiveness of mastectomy vs. breast conserving surgery.

Table 30

Impact of necrosis on the effectiveness of mastectomy vs. breast conserving surgery.

Table 31. Association between treatment and patient outcomes, stratified by microinvasion status.

Table 31

Association between treatment and patient outcomes, stratified by microinvasion status.

Tamoxifen

The NSABP-24 assessed the value of tamoxifen following DCIS diagnosis and found tamoxifen use to reduce risk of recurrent DCIS or invasive carcinoma. The trial found that tamoxifen was associated with a 50 percent reduction in contralateral disease and of breast cancer mortality but had no impact on all-cause mortality (Table 32). Adverse events associated with tamoxifen are consistent with its profile in other settings. There was an increase in hot flushes, fluid retention, and vaginal discharge associated with chemotherapy (Table 33).324 Combined treatment (lumpectomy, radiation, and tamoxifen) compared to lumpectomy and tamoxifen reduced the rates of all cancer events by 29 percent (pooled RR 0.71, 95 percent CI 0.62; 0.82, I squared 0 percent).323,324 The study did not show any differential impact of tamoxifen for women with or without adverse pathological characteristics except for a nonsignificant indication that tamoxifen was less effective for women without comedo necrosis or with smaller tumors.6

Table 32. Effect of tamoxifen on patient outcomes (results from RCTs).

Table 32

Effect of tamoxifen on patient outcomes (results from RCTs).

Table 33. Adverse events after compared treatments.

Table 33

Adverse events after compared treatments.

The only observational study of tamoxifen use after DCIS that included comparisons with nonusers was conducted by Warren.316 They found that women with DCIS who received tamoxifen had the same hazard of local DCIS or invasive cancer as women who did not receive tamoxifen.

Ongoing studies such as the NSABP-37 are examining the comparative effectiveness of tamoxifen and aromitase inhibitors and the use of trastuzumap for Her2 positive women (NSABP B-43).

APBI

An emerging controversy is whether APBI therapy is as effective as whole breast radiation therapy. Observational studies reporting results of APBI for DCIS are limited to the MammoSite® technology, and do not include control groups. Multiple publications about the effectiveness of the MammoSite® technology for DCIS are available (Appendix Table F45). The ongoing NSABP-39 trial randomizes women to whole or APBI therapy.397 For that trial, three partial breast techniques are treated as equivalent: multicatheter brachytherapy, MammoSite® balloon catheter, and 3-D conformational external beam radiation. Other ongoing trials are comparing whole breast to specific types of APBI.

Summary

Randomized trials provide consistent evidence that DCIS treated with breast conserving therapy plus radiation compared to breast conserving therapy alone results in reduced total local recurrence by 53 percent (pooled RR 0.47, 95 percent CI 0.34; 0.63)295,323,324,331 and local invasive breast cancer recurrence by 46 percent (pooled RR 0.54, 95 percent CI 0.43; 0.68)295,323,324,331 with no differences in overall and breast cancer mortality, all295,323,324 or invasive323,324,331 contralateral breast cancer, total distant,323,295,331,398 or local regional nodes recurrence (Table 34).398,399 Observational studies point to somewhat inconsistent effects regarding the benefit of BCS with RT relative to BCS alone. The observational studies, however, are frequently under-powered, subject to selection bias (that is, patients are not randomly allocated to RT or not) and inconsistent in their control of known confounding factors.

Table 34. Summary evidence map: Patient outcomes across treatments.

Table 34

Summary evidence map: Patient outcomes across treatments.

While not studied in a randomized fashion, studies point to equivalent outcomes between breast conserving surgery plus radiation and mastectomy while BCS alone tends to be inferior to mastectomy.

Subset analyses, while generally lower level of evidence (e.g., not always multivariate adjusted) do not point to differential effectiveness of surgery or radiation in the presence of adverse prognostic factors. This lack of differential effect suggests that treatment alone may not eliminate the adverse prognosis but also suggests that for patients with adverse prognostic features, treatment may be particularly important.

Evidence of the effectiveness of tamoxifen for treating DCIS is based on a very small number of randomized and observational studies but is quite promising. Ongoing studies evaluating the value of hormonal therapies and herceptin for use with DCIS will help clarify the benefit of these therapies, particularly if assessment of estrogen and progesterone receptor status and Her2 positivity in the general population increases.

Synthesizing across studies, we found no effects on overall mortality or breast cancer mortality (Table 35). Only one observational study reported significant reduction in crude odds of breast cancer mortality after adjuvant radiotherapy (LR or LRT versus L or LT).316 All cancer events were reduced after combined treatment (lumpectomy plus radio- and chemotherapy) when compared to dual therapy (lumpectomy plus radiotherapy324 or lumpectomy plus tamoxifen).323,324 However, given the low level of mortality associated with DCIS and the long treatment horizon, it is likely that even the largest of these studies is underpowered to identify a mortality benefit. A similar conclusion was reached with invasive breast cancer where mortality is much more common. Yet, until all studies were pooled using meta-analysis, no mortality effect was observed when comparing BCS+RT to BCS alone.

Table 35. Summary evidence map: All cancer events, overall and breast cancer mortality, and adverse events across treatments.

Table 35

Summary evidence map: All cancer events, overall and breast cancer mortality, and adverse events across treatments.

The overall evidence of treatment effectiveness is consistent with treatment effectiveness for invasive breast cancer. This insight should facilitate transfer knowledge about treatment effectiveness from invasive breast cancer to DCIS.