Division of Cancer Control and Population Sciences, National Cancer Institute, Bethesda, Maryland
Health Services and Economics Branch/Applied Research Program, Division of Cancer Control and Population Sciences, National Cancer Institute, Executive Plaza North, Room 4005, 6130 Executive Blvd., MSC 7344, Bethesda, MD 20892-7344
The diagnosis of ductal carcinoma in situ (DCIS) is increasing, although to the authors' knowledge there is no consensus regarding optimal treatment. This analysis of women treated with breast-conserving surgery (BCS) evaluated the impact of radiation therapy (RT) in patient outcomes.
The current study included a population-based sample of 1103 women residing in selected Surveillance, Epidemiology, and End Results (SEER) registries who were diagnosed with DCIS between 1991–1992. Data were obtained from the registry, physician follow-up, and pathology reports. Physicians were contacted in 1999 to determine whether the patient had developed a second event in the ipsilateral breast. For second events, pathology reports were reviewed to determine the presence of in situ or invasive disease. Registry data through 2001 were used to assess death rates and cause of death. Cox proportional hazards and logistic regression models were used to evaluate the rates of second events and breast carcinoma deaths between women treated with and without RT.
Over an average of 91 months, 13.2% of women developed a second event. Rates of second events were 11% for women treated with BCS and RT compared with 15% for women treated with BCS only (adjusted hazards ratio, 0.64; 95% confidence interval, 0.44–0.92). Women receiving RT were significantly less likely to develop invasive breast carcinoma in the ipsilateral breast (adjusted odds ratio, 0.40). By 2001, the rate of death from breast carcinoma was 2.7%; in the group of women treated with BCS only compared with 0.8% in the group of women treated with BCS with RT.
Ductal carcinoma in situ (DCIS) is a heterogeneous disease, the natural history of which is not well understood. In recent years there has been a significant increase in the incidence of DCIS, largely attributed to the greater use of screening mammography.1, 2 Increased detection has resulted in more DCIS cases with localized microscopic lesions. The heterogeneity of the condition, the detection of smaller lesions, and a limited understanding of the natural progression of the disease have resulted in patients newly diagnosed with DCIS and their physicians facing uncertainty regarding what is adequate, but not excessive, treatment.
Historically, DCIS has been treated with total mastectomy, with a risk of recurrence of < 1%. More recently, breast-conserving surgery (BCS) has been offered as an option to women. Based on clinical trials and case series of women with DCIS undergoing BCS, approximately 10–20% will develop a recurrence within 5 years; approximately half of these recurrences will be with invasive disease.3–5 Clinical trials, including the National Surgical Adjuvant Breast and Bowel Project (NSABP), European Organization for Research and Treatment of Cancer (EORTC), and United Kingdom (U.K.) trials, have concluded that for women undergoing BCS for DCIS, the addition of radiation therapy (RT) after surgery reduces the risk of recurrence.3, 4, 6, 7 However, clinical trials do not always include an array of cases representative of those in the general population8 and the efficacy of a treatment within a controlled setting may not reflect the outcomes of treatment provided in the community setting.9 To our knowledge, to date only one population-based study has evaluated the risk of a second tumor in the ipsilateral breast after a DCIS diagnosis by treatment modality.5
The purpose of this analysis is to provide an assessment of factors associated with an increased risk of developing a second tumor in the ipsilateral breast among a population-based cohort of women with DCIS who were treated with BCS. The primary focus was to evaluate the effectiveness of RT in preventing a second tumor. In addition, for women who had a second tumor, we assessed the risk of developing invasive or in situ disease as well as the risk of death from breast carcinoma.
MATERIALS AND METHODS
Data for this analysis were obtained from multiple sources. Women diagnosed with DCIS were identified from the National Cancer Institute's (NCI) Surveillance, Epidemiology, and End Results (SEER) program data. The SEER program is comprised of population-based cancer registries that obtain data for all incident cancer cases occurring within a defined geographic area. The cancer registrars collect data regarding tumor characteristics and initial treatment as well as patient demographic information. This information is obtained primarily through the hospital medical records, outpatient surgical centers, or pathology departments.
With many cancer patients receiving treatment outside the hospital setting, the routine SEER data regarding the first course of treatment may be incomplete. Therefore, the NCI sponsors the Patterns of Care (POC) Studies to obtain details of therapy not found in the hospital record. As part of the POC Studies, selected SEER registries undertake special data collection efforts that include reabstraction of the medical record and contacting the treating physician for more in-depth information regarding the characteristics of the cancer, information concerning the treatment provided, and any comorbid conditions.10 Registries from Atlanta, Connecticut, Detroit, Iowa, Los Angeles, New Mexico, San Francisco, San Jose, and Seattle participated in the DCIS study and provided POC data.
For this project, data were collected from the physician and the medical record for two different time periods. The first data collection period related to the initial treatment and any patient comorbidities that were present at the time each woman was diagnosed with DCIS in 1991 or 1992. The second data collection was to obtain information regarding whether second tumors in the ipsilateral breast, either recurrent or new breast primary tumors, had occurred between the time of initial diagnosis to the end of 1999. For women with a second tumor, data were collected from the physician and the medical record to determine when the second tumor had occurred.
In addition to the SEER and POC data, more detailed clinical and histologic information from the time of the initial diagnosis was needed to evaluate factors that were potentially prognostic for developing a second tumor after a DCIS diagnosis. Therefore, we obtained pathology reports from the time of diagnosis for any biopsy or BCS for all DCIS cases. We also obtained the pathology reports for subsequent tumors developing in the same breast.
The sampling frame for the current study included women diagnosed with DCIS from the SEER registries participating in this study. Women were diagnosed with DCIS in 1991 or 1992, with the exception of cases from Los Angeles County, which were only diagnosed in 1992 because this registry did not join the SEER program until that year. DCIS was defined based on the American Joint Committee on Cancer (AJCC) definition (behavior code of in situ and histology codes of 8500, 8501, 8503, 8521, and 8522).11 Women with a previous diagnosis of cancer (except for nonmelanoma skin cancers) were excluded from the study, as were women with simultaneous cancer diagnoses, because this might have altered the selection of therapy. We included only women who were treated with BCS (SEER site specific surgery code 10–28).12
Cases for the current study were randomly selected from women diagnosed with DCIS in each registry. Women were stratified by age, which served as a proxy for premenopausal and postmenopausal status Women age 50 years and younger were classified as premenopausal whereas women age 51 years and older were considered to be postmenopausal. Within each strata, potential cases also were stratified based on whether a woman received RT after surgery. To obtain more stable estimates, younger women and women receiving RT were oversampled. The sample included 50% of all breast carcinoma patients treated with BCS plus RT and 46% of patients who underwent BCS only.
Patient demographic characteristics included in the current analysis were obtained from the SEER data. These included age (age younger than 51 yrs, ages 51–64 yrs, and age 65 yrs and older), race (white, black, or other), and marital status (married or unmarried [including single, separated/divorced, or widowed]). Comorbid conditions that were present at the time of the initial diagnosis of DCIS were recorded as International Classification of Diseases, 9th revision (ICD-9) diagnoses on the POC data. We used the Deyo adaptation of the Charlson comorbidity score13, 14 to calculate comorbidity weights (0 or 1+). Information regarding surgery and RT at the time of diagnosis was obtained from the SEER data or, if the information was not available from the SEER file, the POC data then were used. RT was defined from the SEER data as the administration of external beam RT.
Prognostic Factors from the Pathology Reports
Prior to reviewing the pathology reports, the study pathologists (D.W., T.B., and C.K.) identified five prognostic factors to be included in the analysis. These factors were generally agreed on as significant in the period between 1991–199215 and included the size of the tumor, status of the surgical excision margins, the presence of comedocarcinoma, the presence of necrosis, and overall grade of the DCIS. These factors were categorized as follows.
Tumor size was recorded as less than 1 cm, 1 cm to less than 2 cm, and 2 cm or greater. Size was recorded as stated in the report or, if not stated, was an imputed value determined from the number of microscopic sections involved, the size of the biopsy specimen, and the presence or absence of residual disease if there was a reexcision performed. The status of surgical excision margins (involved or clear) was taken directly from the report. If there was no mention of the margins on the pathology report, then surgical margin status was defined as not reported. Comedocarcinoma was recorded on the basis of a specific statement or classic description in the report. If necrosis was noted in the report but not specifically designated as comedonecrosis, it was recorded as necrosis. We found the presence of comedocarcinoma to be highly correlated with the overall grade of the tumor. A previous study by Lagios and Silverstein suggested that nuclear grade, rather than the presence of comedocarcinoma, is a more important predictor of recurrence16; therefore, comedocarcinoma was excluded from the final analysis. The overall grade of the tumor was recorded directly from the report if stated. When it was not expressly stated, tumor grade was estimated based on any description of the nuclear features (if given), including nuclear size, chromatin pattern, nucleoli, and mitotic activity; atypia; architectural pattern of the DCIS; and the presence or absence of necrosis. Nuclear features, when described, were given more weight than architectural features. Features such as bizarre mitoses or large prominent nucleoli were considered indicators of high-grade DCIS. Interpretation of descriptions was guided by reviewing published texts available between 1991 and 1992.15 The assignment of overall grade and nuclear grade was also aided by review of descriptive material on the Radiation Therapy Oncology Cooperative Group website,17 which defines grades of DCIS and provides case examples. Overall grade was initially classified as low (scored as 1), intermediate (scored as 2), or high (scored as 3). Because the low-grade and intermediate-grade lesions had similar outcomes, they subsequently were combined for statistical analysis.
Two pathologists independently reviewed all pathology reports initially, with each pathologist then reviewing approximately two-thirds of the reports. In cases in which the two pathologists did not agree on one or more of the specific features noted earlier, the third pathologist independently reviewed the report. If the third opinion agreed with one of the other two, the majority interpretation was accepted as the final one. If the third opinion differed from both of the original two, the case was discussed among the three and a consensus interpretation was reached. In the occasional case in which no consensus was reached, the case was excluded from the statistical analysis. The absence of information regarding one or more parameters also was recorded and cases with more than one factor missing on the report were excluded from the analysis.
There were three outcomes included in the current analysis: occurrence of a second tumor in the ipsilateral breast, either recurrence or a new primary tumor; whether this event was invasive or in situ; and whether the woman died of breast carcinoma. Second tumors in the ipsilateral breast were identified from the POC data as either a recurrence or a new primary tumor or from a new primary tumor reported in the SEER data. The month and year of the second tumor was extracted from the reporting source.
For women who experienced a second tumor, the pathology reports for the later event were independently reviewed by two pathologists to determine whether the second tumor was invasive or in situ. In the event there was disagreement regarding the second tumor, the report was reviewed by a third pathologist. In the few cases in which the pathology report was not available, we used the behavior code as reported on the SEER data or the follow-up POC data. Breast carcinoma-related deaths were identified from the information provided on the SEER data. The SEER registries perform follow-up of cases in their files by linking to the state death certificates to obtain the vital status and cause of death information listed on the death certificate. For our analysis, vital status information was available through 2001. We identified all cases within our sample with the cause of death reported as breast carcinoma. Because of concerns regarding the potential misclassification of breast carcinoma deaths as death due to other cancers, we included deaths due to common metastatic sites for breast carcinoma –lung, liver, bone, or brain– unless the women had a SEER record with a report of a second primary tumor coded as lung, liver, bone, brain, or contralateral breast.
The Cox proportional hazards model was used to determine the relative risks (RRs) for a second tumor developing between the time of diagnosis and 1999, adjusting for various demographic and clinical factors. We used a multivariate logistic regression model to calculate the odds ratios (ORs) for factors that may be associated with the event being invasive. The model contained the same covariates as did the Cox proportional hazards model. Adjusted survival curves were presented comparing rates of recurrence over time for selected risk factors (RT, tumor size, and overall grade of the tumor).
In the current sample, there were a total of 1103 women diagnosed with DCIS in 1991 and 1992 who were treated with BCS. Among these women, 477 (43.2%) also received RT. As can be seen in Table 1, women who received BCS and RT were younger and had larger sized tumors, more necrosis, and higher grade lesions compared with women undergoing BCS only. Women treated with BCS and RT also were more likely to receive tamoxifen than women treated with BCS only (22.6% vs. 9.7%).
Table 1. Characteristics of a Cohort of Women Treated with BCS after a Diagnosis of DCIS in 1991–1992
Chi-square calculations are based on a comparison of the patients treated with breast-conserving surgery (BCS) only with the patients treated with BCS plus radiation therapy.
Age group in yrs
Charlson comorbidity score
< 1 cm
1–< 2 cm
≥ 2+ cm
Clear surgical margins
Overall tumor grade
Women in the sample were followed for a mean of 91 months. During the period of follow-up, 13.2% of the women (n = 145) developed a second occurrence of carcinoma in the ipsilateral breast. For women treated with BCS only, this percentage was 15% (94 of 626 patients) compared with 10.7% (51 of 477 patients) for women treated with BCS and RT—a 29% reduction in the risk of second tumors. In a Cox model, the adjusted hazards ratio (HR) for all second events was 0.64 for women undergoing RT, which is significantly lower than that for women who had received no RT (Table 2). Several clinical and demographic factors were found to be significantly associated with an increased risk of a second tumor. These included the overall grade of the DCIS (HR, 1.76) and cases in which the surgical margin status was not reported on the pathology report (HR, 1.96). Patients who were unmarried (HR, 1.52) and those with a Charlson comorbidity score of 1 or greater (HR, 1.62) were more likely to develop a second tumor.
Table 2. Assessment of the Risk of Any Second Breast Tumor in the Ipsilateral Breast and the Odds of an Invasive and Noninvasive Second Tumor for Women Diagnosed with DCIS and Treated with BCS
Hazards ratios were calculated from a Cox survival model. The odds ratios were calculated from a logistic regression model.
< 1 cm
1–< 2 cm
≥ 2+ cm
Clear surgical margins
Overall tumor grade
Age group in yrs
Charlson comorbidity scores
Of those women with a second event, 43% (n = 62) developed an invasive tumor (31% for women treated with BCS and RT vs. 49% for women treated with BCS only). Receiving RT was found to reduce a woman's chance of developing an invasive tumor by approximately one-third (37%). In the adjusted model, among women with a second event, those who were treated with BCS plus RT were significantly less likely to develop invasive breast carcinoma in the ipsilateral breast (OR, 0.40), although RT did not significantly decrease the risk of noninvasive disease. Women who were unmarried or whose surgical margin status was not reported were at a significantly greater risk of having invasive disease with a second tumor (OR of 2.07 and OR of 1.93, respectively), but were not significantly more likely to have noninvasive events. Overall grade was found to be associated with an increased risk of a second tumor and of noninvasive tumors. Overall grade was not found to be significantly associated with an increased risk of invasive disease.
The figures show the adjusted curves for event-free survival by month through the end of 1999. Figure 1 compares the event rate between women who did and women who did not receive RT. Figures 2 and 3 show survival curves by tumor size and overall grade, respectively. For women with tumor sizes measuring less than 2 cm, the event-free survival at 100 months was 86%, compared with 79% for women with a tumor measuring larger than 2 cm. The difference in event-free survival was even greater based on overall grade. The adjusted model predicted 86% event-free survival at 100 months among women with low-grade and intermediate-grade DCIS compared with 77% for women with high-grade lesions.
Twenty-two breast carcinoma deaths had occurred by the end of 2001. These deaths occurred in 2.0% of the women (Table 3). Death was more frequent in women treated with BCS only (2.7%) compared with women treated with both BCS and RT (0.8%) (P = 0.02, unadjusted), although the patients treated with BCS plus RT tended to have worse tumor grade and a larger tumor size.
Table 3. Number and Percentage of Breast Carcinoma Related Deaths through 2001 after a Diagnosis of DCIS in 1991–1992 among Women Treated with BCS, with and without RT
The current analysis used population–based data for women with DCIS to enhance the current understanding of factors that are associated with an increased risk of developing a second tumor in the ipsilateral breast. Women who underwent BCS followed by RT were found to have a significantly reduced risk of developing a second tumor compared with women who did not receive RT (11% vs. 15%). Our finding that RT is protective is consistent with reports from clinical trials.3, 6, 7 The benefit of RT varied by trial, with the NSABP and U.K. trials reporting 5-year rates of second events to be 16% for women treated with BCS only compared with 7% for women treated with BCS and RT.3, 7 The EORTC study found at 4 years that 16% of the women treated with BCS had experienced only a local recurrence compared with a rate of 9% for women undergoing BCS with RT.6 Observational studies reported similar findings, with 5-year recurrence rates ranging from 6–25% in those patients treated with BCS only whereas women who received BCS with RT had reported rates of recurrence rates of 0–14% in.5, 18
The clinical trials and the majority of observational studies to date have reported that the use of RT reduces the risk of a second event by approximately half. The current study found the benefit of RT to be less than that reported in these studies. The greater benefit from RT reported in the earlier studies, many of which were initiated before the widespread adaptation of mammography, may reflect outcomes for cases that were detected clinically versus mammographically. The lower benefit of RT that we observed also may reflect that fact that in our observational study, patients were not randomly assigned to treatment. Women in the RT group had more known risk factors for recurrence (tumor size, overall tumor grade, and necrosis) compared with women in the non-RT group. We used a multivariate Cox proportional hazards model to adjust for these differences. However, there may be unmeasured differences between these two groups that contributed to the number of second events.
For women who developed a second tumor, we found that patients who did not receive RT were significantly more likely to develop a subsequent tumor that was invasive. This finding is similar to results from another population-based study5 as well as earlier clinical trials.3, 6, 7 The development of an invasive second event after DCIS has serious ramifications for patient outcomes. A recent case series study reported that of those women initially diagnosed with DCIS, 49% had AJCC Stage II or worse disease when the recurrence was invasive carcinoma.19
DCIS, in the absence of invasive recurrence, is considered to be a disease with little risk of mortality. In the current study, only 2.0% of women had a breast carcinoma-related death between their DCIS diagnosis in the period between 1991 and 1992 and the end of nearly 10 years of follow-up in 2001. This death rate is similar to the rate of 1–2% reported from clinical trials3, 6, 7 and from an earlier study using SEER data, in which 0.7% of women diagnosed in the postmammography era (1984–1989) died of breast carcinoma within 5 years.20
We did observe a higher rate of death from breast carcinoma for women who did not receive RT compared with women who were treated with RT, although we were not able to adjust for differences between the two groups because of the small number of events. Although this unadjusted difference reached statistical significance, the number of events was sufficiently small that any conclusions regarding the benefit of RT on cancer survival should be viewed with caution. However, these findings do suggest that the risk of death from breast carcinoma cannot be dismissed and should be monitored in future analyses of DCIS outcomes.
To our knowledge to date, much of the data regarding survival after DCIS treatment comes from clinical trials, in which patient follow-up, including regular mammography, is closely monitored. Population-based studies include patients whose follow-up may be less optimal than in clinical trials. Studies using population-based data have reported that women with breast carcinoma who were treated with BCS exclusive of RT had lower rates of mammographic surveillance than did women treated with BCS and RT.21, 22 The underuse of mammography is more common among unmarried women,23 which may in part explain our finding that unmarried women had higher rates of invasive second events. However, unmarried women had comparable rates of second DCIS tumors, suggesting that some of the unmarried group were receiving postdiagnostic mammograms. Unfortunately, we did not have information regarding mammography use to evaluate its role in detecting subsequent breast tumors.
To our knowledge, there have been many studies that have assessed clinical factors as prognostic of a DCIS recurrence, although to date there is no consensus regarding which factors are most important. High tumor grade, increased tumor size, and necrosis have been frequently cited as risk factors for disease recurrence.24 We found that high overall grade was a statistically significant prognostic factor. A tumor size greater than 2 cm demonstrated a moderate association with disease recurrence, but did not achieve statistical significance. However, tumor size is difficult to assess retrospectively and should not be dismissed as unimportant. Earlier studies also have included patient age and surgical margin status as potential prognostic factors.25, 26 We found no association between patient age and the risk of a second event. With regard to surgical margin status, we found that patients with undetermined margin status were significantly more likely to develop a second tumor compared with patients in whom clear surgical margins were reported. A recent study by Kerlikowske et al. reached a similar finding.27 One likely explanation for the increased risk of a second event in women whose surgical margin status was undetermined is that both of these studies used cases from a time before there was widespread agreement regarding the need for clear surgical margins and the definition of acceptable margin width. It is likely that in cases in which the surgical margin status could not be determined from review of the pathology report, the pathologist did not understand, at that time period, the importance of surgical margin status and that the initial pathologic examination was inadequate for an evaluation of the surgical margins. Therefore, the likelihood of complete excision would be lower than for a woman in whom a more concerted effort was made to determine the surgical margin status.
The current study had several limitations. This study is a retrospective analysis using observational data. Clinical trials are considered the standard by which the utility of treatments is measured. Nevertheless, outcomes from DCIS trials are influenced not only by the randomized treatment but also by the uniformity of treatment, including standardized RT administration and routine postdiagnostic surveillance. Therefore, trial results may not be completely applicable to women in the community setting. We do not have data that would be available from a clinical trial, including the dose, boost, and volume of RT received by women in our sample. An additional limitation of the current study is that the majority of the prognostic factors were obtained from a retrospective review of the pathology report. There is always some concern regarding whether the pathology report accurately reflects what was on the slides. To address this concern, we conducted a validation of the pathology report against the slides for a subset of 183 cases.28 We found that the degree to which the pathology reports matched the slides varied by prognostic factors. The positive predictive value of the prognostic factors ranged from 80–97%.
Among women diagnosed with DCIS and treated in the community, RT was found to significantly reduce the risk of a second event in the ipsilateral breast. Moreover, RT reduced the chance that second events was invasive. However, the use of RT is not a benign treatment and there currently is significant interest in identifying which prognostic factors can be used to identify those women with a low risk of disease recurrence who can safely forego RT after BCS.
The use of prognostic factors is one aspect of deciding whether to use RT. In addition, compliance with postsurgical care, such as RT and regular mammography surveillance to detect disease recurrence as early as possible, must be evaluated as well. Earlier studies have demonstrated that among those women treated in the community, there is less than optimal use of mammography after a diagnosis of breast carcinoma, especially for women treated with BCS alone. Therefore, the decision regarding the optimal treatment of DCIS in the community must consider not only prognostic clinical factors but also the need for regular postdiagnostic surveillance and consideration of patient compliance with follow-up care.
The authors acknowledge the programming efforts of Joe Zou and James Cucinelli, Information Management Services Inc.