[Skip to Content]
Sign In
Individual Sign In
Create an Account
Institutional Sign In
OpenAthens Shibboleth
[Skip to Content Landing]
Download PDF
Figure 1.
Patient Population
Patient Population

SEER indicates Surveillance, Epidemiology, and End Results.

Figure 2.
Kaplan-Meier Curves for Breast Cancer–Specific Survival Among a Propensity-Weighted Cohort of Patients
Kaplan-Meier Curves for Breast Cancer–Specific Survival Among a Propensity-Weighted Cohort of Patients
Table 1.  
Patient Characteristics by Receipt of Cancer-Directed Surgery
Patient Characteristics by Receipt of Cancer-Directed Surgery
Table 2.  
Effect of Surgery on BCSS According to Nuclear Grade
Effect of Surgery on BCSS According to Nuclear Grade
Table 3.  
Effect of Surgery on OS According to Nuclear Grade
Effect of Surgery on OS According to Nuclear Grade
1.
Howlader  N, Noone  A, Krapcho  M,  et al. National Cancer Institute. SEER Cancer Statistics Review: 1975-2011. http://seer.cancer.gov/csr/1975_2012/. Accessed April 29, 2015.
2.
Allegra  CJ, Aberle  DR, Ganschow  P,  et al.  National Institutes of Health State-of-the-Science Conference statement: diagnosis and management of ductal carcinoma in situ September 22-24, 2009. J Natl Cancer Inst. 2010;102(3):161-169.
PubMedArticle
3.
Allred  DC, Mohsin  SK, Fuqua  SA.  Histological and biological evolution of human premalignant breast disease. Endocr Relat Cancer. 2001;8(1):47-61.
PubMedArticle
4.
Welch  HG, Black  WC.  Using autopsy series to estimate the disease “reservoir” for ductal carcinoma in situ of the breast: how much more breast cancer can we find? Ann Intern Med. 1997;127(11):1023-1028.
PubMedArticle
5.
Page  DL, Dupont  WD, Rogers  LW, Jensen  RA, Schuyler  PA.  Continued local recurrence of carcinoma 15-25 years after a diagnosis of low grade ductal carcinoma in situ of the breast treated only by biopsy. Cancer. 1995;76(7):1197-1200.
PubMedArticle
6.
Page  DL, Dupont  WD, Rogers  LW, Landenberger  M.  Intraductal carcinoma of the breast: follow-up after biopsy only. Cancer. 1982;49(4):751-758.
PubMedArticle
7.
Ozanne  EM, Shieh  Y, Barnes  J, Bouzan  C, Hwang  ES, Esserman  LJ.  Characterizing the impact of 25 years of DCIS treatment. Breast Cancer Res Treat. 2011;129(1):165-173.
PubMedArticle
8.
Benson  JR, Wishart  GC.  Predictors of recurrence for ductal carcinoma in situ after breast-conserving surgery. Lancet Oncol. 2013;14(9):e348-e357.
PubMedArticle
9.
Wapnir  IL, Dignam  JJ, Fisher  B,  et al.  Long-term outcomes of invasive ipsilateral breast tumor recurrences after lumpectomy in NSABP B-17 and B-24 randomized clinical trials for DCIS. J Natl Cancer Inst. 2011;103(6):478-488.
PubMedArticle
10.
Wang  S-Y, Shamliyan  T, Virnig  BA, Kane  R.  Tumor characteristics as predictors of local recurrence after treatment of ductal carcinoma in situ: a meta-analysis. Breast Cancer Res Treat. 2011;127(1):1-14.
PubMedArticle
11.
Virnig  BA, Tuttle  TM, Shamliyan  T, Kane  RL.  Ductal carcinoma in situ of the breast: a systematic review of incidence, treatment, and outcomes. J Natl Cancer Inst. 2010;102(3):170-178.
PubMedArticle
12.
Allegra  CJ, Aberle  DR, Ganschow  P,  et al.  NIH state-of-the-science conference statement: diagnosis and management of ductal carcinoma in situ (DCIS). NIH Consens State Sci Statements. 2009;26(2):1-27.
PubMed
13.
Little  RJA, Rubin  DB. Statistical Analysis With Missing Data. 2nd ed. New York, NY: Wiley; 2002.
14.
Horton  NJ, Kleinman  KP.  Much ado about nothing: a comparison of missing data methods and software to fit incomplete data regression models. Am Stat. 2007;61(1):79-90.
PubMedArticle
15.
Royston  P, White  IR.  Multiple imputation by chained equations (MICE): implementation in Stata. J Stat Softw. 2011;45:1-20.Article
16.
van Buuren  S, Boshuizen  HC, Knook  DL.  Multiple imputation of missing blood pressure covariates in survival analysis. Stat Med. 1999;18(6):681-694.
PubMedArticle
17.
Rosenbaum  PR.  Model-based direct adjustment. J Am Stat Assoc. 1987;82:387-394.Article
18.
Austin  PC.  An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivariate Behav Res. 2011;46(3):399-424.
PubMedArticle
19.
Patorno  E, Grotta  A, Bellocco  R, Schneeweiss  S.  Propensity score methodology for confounding control in health care utilization databases. Epidemiol Biostat Public Health. 2013;10(3).
20.
Rosenbaum  PR, Rubin  DB.  Reducing bias in observational studies using subclassification on the propensity score. J Am Stat Assoc. 1984;79:516-524.Article
21.
Esserman  L, Shieh  Y, Thompson  I.  Rethinking screening for breast cancer and prostate cancer. JAMA. 2009;302(15):1685-1692.
PubMedArticle
22.
Esserman  LJ, Thompson  IM, Reid  B,  et al.  Addressing overdiagnosis and overtreatment in cancer: a prescription for change. Lancet Oncol. 2014;15(6):e234-e242.
PubMedArticle
23.
Welch  HG, Black  WC.  Overdiagnosis in cancer. J Natl Cancer Inst. 2010;102(9):605-613.
PubMedArticle
24.
Bleyer  A, Welch  HG.  Effect of three decades of screening mammography on breast-cancer incidence. N Engl J Med. 2012;367(21):1998-2005.
PubMedArticle
25.
Wilt  TJ, Brawer  MK, Jones  KM,  et al; Prostate Cancer Intervention vs Observation Trial (PIVOT) Study Group.  Radical prostatectomy versus observation for localized prostate cancer. N Engl J Med. 2012;367(3):203-213.
PubMedArticle
26.
Kerlikowske  K, Molinaro  AM, Gauthier  ML,  et al.  Biomarker expression and risk of subsequent tumors after initial ductal carcinoma in situ diagnosis. J Natl Cancer Inst. 2010;102(9):627-637.
PubMedArticle
27.
Lazzeroni  M, Guerrieri-Gonzaga  A, Botteri  E,  et al.  Tailoring treatment for ductal intraepithelial neoplasia of the breast according to Ki-67 and molecular phenotype. Br J Cancer. 2013;108(8):1593-1601.
PubMedArticle
28.
Vandenbussche  CJ, Khouri  N, Sbaity  E,  et al.  Borderline atypical ductal hyperplasia/low-grade ductal carcinoma in situ on breast needle core biopsy should be managed conservatively. Am J Surg Pathol. 2013;37(6):913-923.
PubMedArticle
29.
Bartlett  JMS, Nofech-Moses  S, Rakovitch  E.  Ductal carcinoma in situ of the breast: can biomarkers improve current management? Clin Chem. 2014;60(1):60-67.
PubMedArticle
30.
Siziopikou  KP, Anderson  SJ, Cobleigh  MA,  et al.  Preliminary results of centralized HER2 testing in ductal carcinoma in situ (DCIS): NSABP B-43. Breast Cancer Res Treat. 2013;142(2):415-421.
PubMedArticle
31.
Sterne  JAC, White  IR, Carlin  JB,  et al.  Multiple imputation for missing data in epidemiological and clinical research: potential and pitfalls. BMJ. 2009;338:b2393.
PubMedArticle
32.
Howlader  N, Noone  AM, Yu  M, Cronin  KA.  Use of imputed population-based cancer registry data as a method of accounting for missing information: application to estrogen receptor status for breast cancer. Am J Epidemiol. 2012;176(4):347-356.
PubMedArticle
33.
Cooper  GS, Virnig  B, Klabunde  CN, Schussler  N, Freeman  J, Warren  JL.  Use of SEER-Medicare data for measuring cancer surgery. Med Care. 2002;40(8)(suppl):IV-43-IV-48.
PubMed
34.
Noone  AM, Lund  JL, Mariotto  A,  et al.  Comparison of SEER treatment data with Medicare claims [published online March 15, 2014]. Med Care. 2014.
PubMed
Original Investigation
August 2015

Survival Benefit of Breast Surgery for Low-Grade Ductal Carcinoma In SituA Population-Based Cohort Study

Author Affiliations
  • 1Department of Surgery, Brigham and Women’s Hospital, Boston, Massachusetts
  • 2Department of Surgery, McGill University Health Centre, Montreal, Québec, Canada
  • 3Harvard School of Public Health, Boston, Massachusetts
  • 4Department of Breast Surgery, Cerrahpasa Medical School, Istanbul University, Istanbul, Turkey
  • 5Department of Breast Oncology Center, Dana-Farber Cancer Institute, Boston, Massachusetts
  • 6Department of Biostatistics and Computational Biology, Dana-Farber Cancer Institute, Boston, Massachusetts
JAMA Surg. 2015;150(8):739-745. doi:10.1001/jamasurg.2015.0876
Abstract

Importance  While the prevalence of ductal carcinoma in situ (DCIS) of the breast has increased substantially following the introduction of breast-screening methods, the clinical significance of early detection and treatment for DCIS remains unclear.

Objective  To investigate the survival benefit of breast surgery for low-grade DCIS.

Design, Setting, and Participants  A retrospective longitudinal cohort study using the Surveillance, Epidemiology, and End Results (SEER) database from October 9, 2014, to January 15, 2015, at the Dana-Farber/Brigham and Women’s Cancer Center. Between 1988 and 2011, 57 222 eligible cases of DCIS with known nuclear grade and surgery status were identified.

Exposures  Patients were divided into surgery and nonsurgery groups.

Main Outcomes and Measures  Propensity score weighting was used to balance patient backgrounds between groups. A log-rank test and multivariable Cox proportional hazards model was used to assess factors related to overall and breast cancer–specific survival.

Results  Of 57 222 cases of DCIS identified in this study, 1169 cases (2.0%) were managed without surgery and 56 053 cases (98.0%) were managed with surgery. With a median follow-up of 72 months from diagnosis, there were 576 breast cancer–specific deaths (1.0%). The weighted 10-year breast cancer–specific survival was 93.4% for the nonsurgery group and 98.5% for the surgery group (log-rank test, P < .001). The degree of survival benefit among those managed surgically differed according to nuclear grade (P = .003). For low-grade DCIS, the weighted 10-year breast cancer–specific survival of the nonsurgery group was 98.8% and that of the surgery group was 98.6% (P = .95). Multivariable analysis showed there was no significant difference in the weighted hazard ratios of breast cancer–specific survival between the surgery and nonsurgery groups for low-grade DCIS. The weighted hazard ratios of intermediate- and high-grade DCIS were significantly different (low grade: hazard ratio, 0.85; 95% CI, 0.21-3.52; intermediate grade: hazard ratio, 0.23; 95% CI, 0.14-0.42; and high grade: hazard ratio, 0.15; 95% CI, 0.11-0.23) and similar results were seen for overall survival.

Conclusions and Relevance  The survival benefit of performing breast surgery for low-grade DCIS was lower than that for intermediate- or high-grade DCIS. A prospective clinical trial is warranted to investigate the feasibility of active surveillance for the management of low-grade DCIS.

Introduction

Ductal carcinoma in situ (DCIS) is a breast lesion defined as a proliferation of monoclonal epithelial cells in breast ducts without evidence of invasion in the basement membranes. Because the prevalence of breast cancer screening has increased, early detection has contributed to a dramatic increase in the incidence of DCIS, which has risen from 5.83 per 100 000 women in 1973 to 35.54 per 100 000 women in 2011.1 Assuming constant incidence and survival rates, it is estimated that by 2020, more than 1 million women living in the United States will have a diagnosis of DCIS.2

Ductal carcinoma in situ displays a wide spectrum of histological diversity along a continuum, ranging from very well to very poorly differentiated, and nuclear grade has accurately conveyed this diversity.3 Approximately 25% to 50% of DCIS cases will likely progress to invasive ductal carcinoma.46 Ozanne et al7 established a simulation model to predict the progression rate of DCIS to clinically significant invasive breast cancer. Ozanne et al7 estimated that the rate of progression from DCIS to invasive cancer across a 10-year period is 60% for high-grade DCIS (for patients younger than 45 years with lesions larger than 1 cm) and 16% for low-grade DCIS (for patients older than 45 years with lesions larger than 2.5 cm). After local therapy for DCIS, nuclear grade was a proven predictive factor of ipsilateral breast cancer recurrence in a randomized clinical trial and meta-analysis.810

An optimal strategy for DCIS management would be based on individual risk factors that predict subsequent invasive ductal carcinoma to avoid overtreatment. Although surgical management is the current standard of care for all grades of DCIS, to our knowledge, the survival benefit of surgical resection has not been examined.11,12 Therefore, we investigated the survival benefit conferred by surgical treatment in patients with DCIS using survival data in the Surveillance, Epidemiology, and End Results (SEER) database. We hypothesized that breast cancer–specific survival (BCSS) for patients with low-grade DCIS is independent from surgical treatment at the time of diagnosis.

Methods
Study Design and Data Source

After receiving an exemption from the Partners HealthCare Institutional Review Board, we performed a retrospective longitudinal cohort study using data from the SEER database of the National Cancer Institute, which has incidence and survival data routinely collected from population-based cancer registries. We used the SEER 9 general health service, which includes cases followed up from 1973 to 2011. The SEER 9 includes data from Atlanta, Georgia, Connecticut, Detroit, Michigan, Hawaii, Iowa, New Mexico, San Francisco, California and Oakland, California, Seattle, Washington, and the Puget Sound area, and Utah. Between January 1, 1988, and December 31, 2011, 96 732 patients who were older than 20 years were diagnosed as having DCIS of the breast (Figure 1). Among these patients, we identified 59 789 cases with a nuclear grade that was available based on pathology results. We excluded patients with Paget disease or ductal carcinoma with microinvasion, patients whose treatment pathway (surgery vs no surgery) was unknown, patients who were listed as deceased prior to recommended surgery, patients in whom it remained unclear whether a surgical procedure was performed owing to a discrepancy in coding, patients with unknown race/ethnicity, patients with unknown radiation status, patients with an unknown method of radiotherapy, and patients with isotope only–based radiotherapy.

Assembly of Key Variables

Using the case listing sessions in SEER*Stat software, version 8.1.5, a data table including individual cancer records and patient characteristics was created that was composed of the following variables: patient identification, year of diagnosis, age, race/ethnicity, histology, tumor size, nuclear grade, estrogen receptor (ER), progesterone receptor (PgR), adjusted American Joint Committee on Cancer 6th tumor node metastasis staging classification, surgery type, reason for no cancer-directed surgery, radiation therapy, cause-specific death classification, other cause of death classification, and survival month. The tumor size was primarily measured according to American Joint Committee on Cancer pathologic staging criteria; however, when pathologic data were not available, the tumor size was measured based on physical or imaging examinations. The hormone receptor status was coded as positive, negative, or unknown based on the records of highest value.

We queried the database using both surgery and therapy codes to determine which patients underwent surgical management of their DCIS. The surgery codes 00 and 02 were used to determine patients who did not have definitive therapy. Agreement between surgery codes and the reasons why a patient did not have cancer-directed therapy were confirmed prior to performing our analysis (Figure 1).

Main Outcome Measure

The SEER database defines mortality data based on the International Classification of Diseases, Eighth to Tenth revisions. In this study, the periods of BCSS and overall survival (OS) were calculated by measuring from the date of diagnosis to the last date for which completed vital status data were available (censored on December 31, 2011). The data for deaths were ascertained from central cancer registries or from state, province, and national registries.

Statistical Analysis

Clinicopathologic factors were compared between the surgery groups and nonsurgery groups using Pearson χ2 tests. For the missing values for covariates including tumor size (n = 14 160; 24.7%), ER (n = 26 274; 45.9%), and PgR status (n = 28 186; 49.2%), we applied a multiple imputation procedure using IVEware macro, version 0.2 (http://www.isr.umich.edu/src/smp/ive)13,14 with the following variables: patient age (continuous), year of diagnosis, race/ethnicity (white, black, or other), ER (positive or negative), PgR (positive or negative), surgery (partial mastectomy, total mastectomy, or nonsurgery), nuclear grade, tumor size classification (0.1-0.5 mm, 0.6-10 mm, 11-50 mm, or <51 mm), and radiation therapy. To stabilize results, the procedure was repeated for 10 cycles to produce a single imputed data set (eTable 1 in the Supplement).15,16

Propensity score weighting was then used to balance patient characteristics between the surgery and nonsurgery groups.17,18 A logistic regression model was used to calculate the probability of receiving surgery, baseline characteristics of patient age (categorical, 5-year interval), year of diagnosis (categorical, 5-year interval), race/ethnicity, tumor size classification, nuclear grade, ER, and PgR. From the model, the inverse predicted probability of breast surgery assignment was used to define weights for patients who received surgery (1/probability) and for those who did not receive surgery (1/[1 − probability]). Patient characteristics after propensity score adjustment are shown to be balanced in eTable 2 in the Supplement.

The hazard ratios for the BCSS and OS of patients in the surgery group compared with patients in the nonsurgery group were evaluated using propensity score weights for log-rank tests and Cox regression models. Adjusted hazard ratios were reported from multivariable Cox proportional hazards models with patient age (categorical, 5-year interval), year of diagnosis (categorical, 5-year interval), tumor size classification, radiation therapy, and African American race/ethnicity. An interaction test was performed to examine whether there was a difference in the survival benefit conferred by surgical treatment according to nuclear grade. To assess the consistency of our findings, we conducted the following 2 types of sensitivity analyses: one by excluding variables of tumor size and hormonal receptor status, with the missing data exchanged by multiple imputation, and another by performing a stratified propensity score analysis after excluding beyond the 2.5th and 97.5th percentiles of the propensity score distribution.19,20

All P values presented were from 2-sided tests that used α = .05 to assess statistical significance. Statistical analyses were performed using SAS, version 9.4 (SAS Institute).

Results
Patient Characteristics in Original Data Set

We identified 57 222 patients with DCIS who were eligible for this study based on our predefined inclusion and exclusion criteria. There were 1169 cases (2.0%) managed without surgery and 56 053 cases (98.0%) managed with surgery. Patient characteristics according to surgery status are shown in Table 1. The proportion of elderly patients, patients with a recent year of diagnosis, African American race/ethnicity, and low-grade DCIS was larger for the nonsurgery group compared with the surgery group.

In the surgery group, partial mastectomy was performed in 34 439 patients (61%), mastectomy was performed in 16 334 patients (29%), and the type of surgery was not known in 5280 patients (9.4%). Among the patients who received partial mastectomy, 23 129 patients (67.2%) underwent radiotherapy of the breast. In the nonsurgery group, the reasons why patients did not receive cancer-directed surgery included the following reasons: a physician did not recommend surgery (n = 547; 46.8%), a physician did not recommend surgery owing to other contradicted conditions (n = 29; 1.7%), a physician recommended surgery but it was not performed because the patient refused (n = 115; 9.8%), and a physician recommended surgery but it was not performed owing to unknown reasons (n = 478; 40.9%).

Survival Benefit of Breast Surgery

During a median follow-up period of 72 months from the date of diagnosis (interquartile range, 34-120 months), there were 576 breast cancer–specific deaths (1.0%) and 3652 deaths from other causes (6.4%). Kaplan-Meier curves of BCSS comparing the surgery and nonsurgery groups in the total patient cohort weighted by inverse propensity scores are shown in Figure 2 (see eTable 3 in the Supplement for analyses of unweighted BCSS). The weighted 10-year BCSS of all patients was 98.5% in the surgery group and 93.4% in the nonsurgery group (absolute difference, 5.1%; log-rank test, P < .001). After adjusting for other clinical factors, nuclear grade remained a statistically significant effect modifier for surgery in BCSS (P = .003).

Survival Benefit of Breast Surgery According to Nuclear Grade

Kaplan-Meier curves of BCSS comparing the surgery and the nonsurgery groups according to nuclear grade are shown in Figure 2. The weighted 10-year BCSS of the surgery and nonsurgery groups were 98.6% and 98.8% in low-grade DCIS (absolute difference, –0.2%), 98.6% and 94.6% in intermediate-grade DCIS (absolute difference, 4.0%), and 98.4% and 90.5% in high-grade DCIS (absolute difference, 7.9%), respectively (Table 2). Multivariate analysis showed that while there was no statistical difference in the weighted hazard ratios of BCSS and OS between the surgery and nonsurgery groups for low-grade DCIS, the weighted hazard ratios of intermediate- and high-grade DCIS were significantly different (Table 2 and Table 3). In the sensitivity analyses, consistent results were observed when excluding variables of tumor size and hormonal receptor status and when performing a stratified propensity score analysis.

Discussion

Low-grade DCIS is an indolent lesion of epithelial cells in breast ducts. It typically exhibits a slow growth pattern and, in many cases, never fully develops into a clinically significant invasive carcinoma.47,21,22 Although the natural history of low-grade DCIS is poorly understood, the current standard therapy is either partial mastectomy followed by whole-breast radiation therapy or a total mastectomy. The choice of local treatments does not currently take the biology of DCIS into consideration. As such, DCIS management principles may benefit from incorporating information regarding the biologic behavior of DCIS based on nuclear grade.2124

In our analysis of a large population-based cohort, we saw a statistically significant survival benefit of breast surgery for intermediate- and high-grade cases of DCIS while no significant survival benefit was appreciated for cases of low-grade DCIS. The weighted BCSS curves of both groups were identical and the prognosis for patients treated nonoperatively was favorable (weighted 10-year BCSS, 98.8%) during the median 72-month follow-up period. From these results, we could consider recommending a strategy of nonoperative management with active surveillance similar to that used in the management of prostate cancer.22,25 In addition, by integrating both conventional pathologic and molecular factors, such as human epidermal growth factor receptor 2, prostaglandin-endoperoxide synthase 2, and Ki67, it may be possible to identify subpopulations of low-grade DCIS with more favorable prognoses.2629 The results from ongoing clinical trials (NCT00290745 and NCT01439711) looking at the role of neoadjuvant therapy for DCIS may provide more information that could help identify optimal biology-based DCIS treatment strategies.30

We used multiple imputation to infer missing data pertaining to tumor size and ER status in the multivariable analysis because it characteristically yields less biased results than other methods. We viewed absent variables in the SEER database as being missing at random, meaning that the systematic difference between the missing and observed values could be explained by differences in observed data.13,31,32 Furthermore, to ensure our study had enough power to detect the survival benefit offered by breast surgery, it was necessary for us to include such patient characteristics. A sensitivity analysis performed by excluding variables, which were exchanged using multiple imputations, confirmed the robustness of our results.

Propensity score weighting is an effective way of removing overt indication biases for the purpose of investigating treatment effects seen in observational studies.1720 Using this method, we attempted to balance the background characteristics of patients between the treatment (surgery) and nontreatment groups (eTable 2 in the Supplement). However, without the performance of a randomized clinical trial, we cannot account for unmeasured confounders that may have influenced the prognosis of DCIS. Our findings justify a subsequent prospective study aimed at investigating new approaches for the management of low-grade DCIS. Such a study may include active surveillance with an imaging arm that includes recommendations for delayed surgery only after confirming progression of the disease.

There were limitations to this study. First, we used population-based cancer registries with limited information regarding patient and tumor characteristics. Unavailable confounders, such as surgical margin status, comorbidity, endocrine therapy, and history of screening, could not be adjusted for this data set. We assumed that our results would have been minimally influenced by including information related to endocrine therapy because ER/PgR were not independent prognostic factors of DCIS in this study. Furthermore, the number of cases that did not receive surgery owing to other contraindicated conditions was much smaller than we expected (n = 29; 1.7%). An additional limitation to our study was that data abstraction was done in part by individuals who could have created information bias. However, it has been reported that there is good agreement between Medicare and SEER data with respect to whether or not patients underwent surgery; therefore, we anticipate that the possibility of misclassification was small.33 There were 159 patients (14% of nonoperative group) who received radiation therapy without undergoing breast surgery, which is not a standard treatment for DCIS. However, we believe these data are most likely accurate because we confirmed the concordance between the surgery code and reason of no cancer-directed surgery using the SEER database. Furthermore, the positive predictive values of receiving radiation therapy between the SEER database and Medicare data are as high as 97.5%, as reported by Noone et al.34 In the SEER database, a biopsy that removes only a fragment or portion of the tumor is recorded as a noncancer-directed treatment. Therefore, we consider this is the group that had radiation therapy without cancer-directed surgery.

We sought to determine the prognosis of patients managed nonoperatively at the time of diagnosis. While we cannot conclude definitively based on the results of this study alone that we can avoid breast surgery for low-grade DCIS, our results do suggest that breast surgery performed at or shortly after the time of diagnosis does not significantly affect survival for low-grade DCIS.

To our knowledge, this study is the first to examine the survival benefit of surgical treatment for DCIS. From the SEER database, we identified more than 1000 cases of DCIS with known nuclear grade managed nonoperatively. The large number of nonoperative cases made it possible to investigate the survival benefit of surgery according to the grade of DCIS after adjusting for other clinicopathologic factors.

Conclusions

The survival benefit of breast surgery for low-grade DCIS was lower than that for intermediate- or high-grade DCIS, which raises concern regarding the necessity and benefit of surgery for patients with low-grade DCIS. A prospective clinical trial is warranted to investigate the feasibility of active surveillance for the management of low-grade DCIS.

Back to top
Article Information

Corresponding Author: Yasuaki Sagara, MD, Division of Surgical Oncology, Dana-Farber/Brigham and Women’s Cancer Center, Harvard Medical School, 450 Brookline Ave, Yawkey Ste 1444, Boston, MA 02215 (ysagara@partners.org).

Accepted for Publication: February 27, 2015.

Correction: This article was corrected on June 11, 2015, for errors in the Author Affiliations and Corresponding Author address and on March 23, 2016, for an error in the abstract.

Published Online: June 3, 2015. doi:10.1001/jamasurg.2015.0876.

Author Contributions: Dr Sagara had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Sagara, Mallory, Aydogan, Golshan.

Acquisition, analysis, or interpretation of data: Sagara, Mallory, Wong, DeSantis, Barry, Golshan.

Drafting of the manuscript: Sagara, Barry, Golshan.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Sagara, Barry.

Administrative, technical, or material support: Sagara, Mallory, Aydogan, Golshan.

Study supervision: Barry, Golshan.

Conflict of Interest Disclosures: None reported.

Previous Presentation: This paper was presented at the 2015 Annual Meeting of the American Society of Clinical Oncology; June 1, 2015; Chicago, Illinois.

Additional Contributions: We acknowledge the instructors at the Harvard School of Public Health Clinical Effectiveness program. The protocol of this study was proposed and refined with their support during the summer program. We also acknowledge Hajime Uno, PhD, Division of Population Sciences and Biostatistics and Computational Biology, Dana-Farber Cancer Institute, for giving us insightful advice on statistical methods. He did not receive financial compensation.

References
1.
Howlader  N, Noone  A, Krapcho  M,  et al. National Cancer Institute. SEER Cancer Statistics Review: 1975-2011. http://seer.cancer.gov/csr/1975_2012/. Accessed April 29, 2015.
2.
Allegra  CJ, Aberle  DR, Ganschow  P,  et al.  National Institutes of Health State-of-the-Science Conference statement: diagnosis and management of ductal carcinoma in situ September 22-24, 2009. J Natl Cancer Inst. 2010;102(3):161-169.
PubMedArticle
3.
Allred  DC, Mohsin  SK, Fuqua  SA.  Histological and biological evolution of human premalignant breast disease. Endocr Relat Cancer. 2001;8(1):47-61.
PubMedArticle
4.
Welch  HG, Black  WC.  Using autopsy series to estimate the disease “reservoir” for ductal carcinoma in situ of the breast: how much more breast cancer can we find? Ann Intern Med. 1997;127(11):1023-1028.
PubMedArticle
5.
Page  DL, Dupont  WD, Rogers  LW, Jensen  RA, Schuyler  PA.  Continued local recurrence of carcinoma 15-25 years after a diagnosis of low grade ductal carcinoma in situ of the breast treated only by biopsy. Cancer. 1995;76(7):1197-1200.
PubMedArticle
6.
Page  DL, Dupont  WD, Rogers  LW, Landenberger  M.  Intraductal carcinoma of the breast: follow-up after biopsy only. Cancer. 1982;49(4):751-758.
PubMedArticle
7.
Ozanne  EM, Shieh  Y, Barnes  J, Bouzan  C, Hwang  ES, Esserman  LJ.  Characterizing the impact of 25 years of DCIS treatment. Breast Cancer Res Treat. 2011;129(1):165-173.
PubMedArticle
8.
Benson  JR, Wishart  GC.  Predictors of recurrence for ductal carcinoma in situ after breast-conserving surgery. Lancet Oncol. 2013;14(9):e348-e357.
PubMedArticle
9.
Wapnir  IL, Dignam  JJ, Fisher  B,  et al.  Long-term outcomes of invasive ipsilateral breast tumor recurrences after lumpectomy in NSABP B-17 and B-24 randomized clinical trials for DCIS. J Natl Cancer Inst. 2011;103(6):478-488.
PubMedArticle
10.
Wang  S-Y, Shamliyan  T, Virnig  BA, Kane  R.  Tumor characteristics as predictors of local recurrence after treatment of ductal carcinoma in situ: a meta-analysis. Breast Cancer Res Treat. 2011;127(1):1-14.
PubMedArticle
11.
Virnig  BA, Tuttle  TM, Shamliyan  T, Kane  RL.  Ductal carcinoma in situ of the breast: a systematic review of incidence, treatment, and outcomes. J Natl Cancer Inst. 2010;102(3):170-178.
PubMedArticle
12.
Allegra  CJ, Aberle  DR, Ganschow  P,  et al.  NIH state-of-the-science conference statement: diagnosis and management of ductal carcinoma in situ (DCIS). NIH Consens State Sci Statements. 2009;26(2):1-27.
PubMed
13.
Little  RJA, Rubin  DB. Statistical Analysis With Missing Data. 2nd ed. New York, NY: Wiley; 2002.
14.
Horton  NJ, Kleinman  KP.  Much ado about nothing: a comparison of missing data methods and software to fit incomplete data regression models. Am Stat. 2007;61(1):79-90.
PubMedArticle
15.
Royston  P, White  IR.  Multiple imputation by chained equations (MICE): implementation in Stata. J Stat Softw. 2011;45:1-20.Article
16.
van Buuren  S, Boshuizen  HC, Knook  DL.  Multiple imputation of missing blood pressure covariates in survival analysis. Stat Med. 1999;18(6):681-694.
PubMedArticle
17.
Rosenbaum  PR.  Model-based direct adjustment. J Am Stat Assoc. 1987;82:387-394.Article
18.
Austin  PC.  An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivariate Behav Res. 2011;46(3):399-424.
PubMedArticle
19.
Patorno  E, Grotta  A, Bellocco  R, Schneeweiss  S.  Propensity score methodology for confounding control in health care utilization databases. Epidemiol Biostat Public Health. 2013;10(3).
20.
Rosenbaum  PR, Rubin  DB.  Reducing bias in observational studies using subclassification on the propensity score. J Am Stat Assoc. 1984;79:516-524.Article
21.
Esserman  L, Shieh  Y, Thompson  I.  Rethinking screening for breast cancer and prostate cancer. JAMA. 2009;302(15):1685-1692.
PubMedArticle
22.
Esserman  LJ, Thompson  IM, Reid  B,  et al.  Addressing overdiagnosis and overtreatment in cancer: a prescription for change. Lancet Oncol. 2014;15(6):e234-e242.
PubMedArticle
23.
Welch  HG, Black  WC.  Overdiagnosis in cancer. J Natl Cancer Inst. 2010;102(9):605-613.
PubMedArticle
24.
Bleyer  A, Welch  HG.  Effect of three decades of screening mammography on breast-cancer incidence. N Engl J Med. 2012;367(21):1998-2005.
PubMedArticle
25.
Wilt  TJ, Brawer  MK, Jones  KM,  et al; Prostate Cancer Intervention vs Observation Trial (PIVOT) Study Group.  Radical prostatectomy versus observation for localized prostate cancer. N Engl J Med. 2012;367(3):203-213.
PubMedArticle
26.
Kerlikowske  K, Molinaro  AM, Gauthier  ML,  et al.  Biomarker expression and risk of subsequent tumors after initial ductal carcinoma in situ diagnosis. J Natl Cancer Inst. 2010;102(9):627-637.
PubMedArticle
27.
Lazzeroni  M, Guerrieri-Gonzaga  A, Botteri  E,  et al.  Tailoring treatment for ductal intraepithelial neoplasia of the breast according to Ki-67 and molecular phenotype. Br J Cancer. 2013;108(8):1593-1601.
PubMedArticle
28.
Vandenbussche  CJ, Khouri  N, Sbaity  E,  et al.  Borderline atypical ductal hyperplasia/low-grade ductal carcinoma in situ on breast needle core biopsy should be managed conservatively. Am J Surg Pathol. 2013;37(6):913-923.
PubMedArticle
29.
Bartlett  JMS, Nofech-Moses  S, Rakovitch  E.  Ductal carcinoma in situ of the breast: can biomarkers improve current management? Clin Chem. 2014;60(1):60-67.
PubMedArticle
30.
Siziopikou  KP, Anderson  SJ, Cobleigh  MA,  et al.  Preliminary results of centralized HER2 testing in ductal carcinoma in situ (DCIS): NSABP B-43. Breast Cancer Res Treat. 2013;142(2):415-421.
PubMedArticle
31.
Sterne  JAC, White  IR, Carlin  JB,  et al.  Multiple imputation for missing data in epidemiological and clinical research: potential and pitfalls. BMJ. 2009;338:b2393.
PubMedArticle
32.
Howlader  N, Noone  AM, Yu  M, Cronin  KA.  Use of imputed population-based cancer registry data as a method of accounting for missing information: application to estrogen receptor status for breast cancer. Am J Epidemiol. 2012;176(4):347-356.
PubMedArticle
33.
Cooper  GS, Virnig  B, Klabunde  CN, Schussler  N, Freeman  J, Warren  JL.  Use of SEER-Medicare data for measuring cancer surgery. Med Care. 2002;40(8)(suppl):IV-43-IV-48.
PubMed
34.
Noone  AM, Lund  JL, Mariotto  A,  et al.  Comparison of SEER treatment data with Medicare claims [published online March 15, 2014]. Med Care. 2014.
PubMed
×