Effectiveness of Radiation Therapy in Older Women With Ductal Carcinoma In Situ

Transcription

1 Effectiveness of Therapy in Older Women With Ductal Carcinoma In Situ Benjamin D. Smith, Bruce G. Haffty, Thomas A. Buchholz, Grace L. Smith, Deron H. Galusha, Justin E. Bekelman, Cary P. Gross Background: For women with ductal carcinoma in situ (DCIS), radiation after conservative surgery lowers the risk of recurrence. However, emerging evidence suggests that radiation confers only a marginal absolute benefit for older women with DCIS. In a cohort of older women with DCIS, we sought to determine whether radiation was associated with a clinically significant benefit. Methods: Using the Surveillance, Epidemiology, and End Results (SEER) Medicare database from January 1, 1992, through December 31, 1999, we identified 3409 women aged 66 years or older treated with conservative surgery for DCIS. A proportional hazards model tested whether radiation was associated with a lower risk of a combined outcome, defined as a subsequent ipsilateral in situ or invasive breast cancer reported by SEER and/or a subsequent mastectomy reported by Medicare claims. The 5-year event risk was determined for patients without and with high-risk features, which were defined as at least one of the following: age years, tumor larger than 2.5 cm, comedo histology, and/or high grade. All statistical tests were two-sided. Results: was associated with a lower risk for each component of the combined outcome (hazard ratio = 0.32, 95% confidence interval [CI] = 0.24 to 0.44). For high-risk patients, the 5-year event risk was 13.6% without radiation versus 3.8% with radiation (difference = 9.8%, 95% CI = 6.5 to 13.2; P <.001). For low-risk patients, the 5-year event risk was 8.2% without radiation versus 1.0% with radiation (difference = 7.2%, 95% CI = 3.6 to 10.9; P <.001). Among healthy women aged years, the number needed to treat with radiation to prevent one event in 5 years was 11 for high-risk patients and for low-risk patients. Conclusion: For older women with DCIS, radiation appears to confer a substantial benefit that remains meaningful even among low-risk patients. [J Natl Cancer Inst 2006;98: ] Ductal carcinoma in situ (DCIS) of the breast has become a major public health concern, with the incidence increasing from 4800 in 1983 (1 ) to more than in 2005 (2 ). Although there are potential benefits to early detection (3 ), the increase in DCIS diagnoses is problematic because substantial controversy still exists regarding its clinical management. Specifically, although randomized trials indicate that radiation to the breast after conservative surgery lowers the risk of recurrence (4 6 ), clinicians continue to debate whether radiation is needed for select patient subgroups with favorable disease characteristics (7 10 ). The necessity of radiation is particularly unclear for women 66 years or older. Current practice guidelines recommend radiation after conservative surgery for this patient group ( 11, 12 ), but emerging evidence reveals that older women experience a low risk of recurrence ( ) and that radiation confers only a marginal benefit for the risk of recurrence ( 9 ), suggesting that older women may not need radiation. We therefore used the Surveillance, Epidemiology, and End Results (SEER) Medicare data from January 1, 1992, through December 31, 1999, to explore whether radiation after conser vative surgery was associated with a clinically significant reduction in the risk of a second breast cancer event, defined as 1) subsequent ipsilateral in situ or invasive breast cancer reported by SEER and/or 2) subsequent mastectomy reported by Medicare claims. P ATIENTS AND METHODS Data Source and Study Sample The National Cancer Institute s (NCI s) SEER Medicare database tracks incident malignancies among Medicare beneficiaries who reside in 11 geographic regions representing 14% of the US population (17 ). During the study period from January 1, 1992, through December 31, 1999, 9208 women aged 66 years or older were diagnosed with noninvasive breast cancer at the age of 66 years or older. All cases of breast cancer were pathologically confirmed, none had evidence of spread to the lymph nodes or distant metastasis, and none were bilateral. Patients were excluded for the following reasons: histology not consistent with ductal origin (n = 558), initial surgical treatment with either biopsy examination only or mastectomy (n = 3183), and history of prior malignancy (n = 160), leaving 5510 women who met all clinical entrance criteria (patients could be excluded for more than one reason). Within this clinical group, patients with a second primary cancer diagnosed within 9 months of the index primary were excluded (n = 147) because billing records could not discriminate between procedures performed for the index cancer versus the second cancer. Patients with inadequate Medicare records (339 without Part A and Part B coverage and 1538 without fee-for-service coverage from 12 months before diagnosis to 9 months after diagnosis) were also excluded. Affi liations of authors: Departments of Therapeutic Radiology (BDS, BGH) and Internal Medicine (GLS, DHG, CPG), Yale University School of Medicine, New Haven, CT; Department of Oncology, Memorial Sloan-Kettering Cancer Center, New York, NY (JEB); Department of Oncology, Robert Wood Johnson Medical School, University of Medicine and Dentistry of New Jersey, New Brunswick, NJ (BGH); Department of Oncology, The University of Texas M. D. Anderson Cancer Center, Houston, TX (TAB). Correspondence to: Benjamin D. Smith, MD, 2200 Bergquist Dr., Ste. 1, Lackland AFB, TX ( bensmith@alumni.rice.edu ). DOI: /jnci/djj359 The Author Published by Oxford University Press. All rights reserved. For Permissions, please journals.permissions@oxfordjournals.org ARTICLES Journal of the National Cancer Institute, Vol. 98, No. 18, September 20, 2006

2 Medicare claims do not consistently report laterality of breast surgery. Because we sought to determine whether radiation lowered the risk of subsequent ipsilateral mastectomy, we excluded patients with unknown laterality at initial diagnosis (n = 6) and patients who developed a contralateral breast cancer (n = 159), leaving 3409 patients for the analysis. Outcomes Patients were considered to be at risk for any outcome during the follow-up period, which began 9 months after diagnosis and continued through December 31, The primary outcome a second breast cancer event was defined as 1) a second, pathologically confirmed, ipsilateral in situ breast cancer reported by SEER; 2) a subsequent, pathologically confirmed, ipsilateral invasive breast cancer reported by SEER (18 ) ; and/or 3) a subsequent mastectomy reported by Medicare claims data. The secondary outcome was a repeat breast-conserving surgery procedure reported by Medicare claims data. Breast-conserving surgery performed after the initial treatment period was not included as a primary outcome because it may be performed as a biopsy procedure and is therefore not specific for recurrence. Covariates In accordance with prior studies, the initial treatment period was defined as the first 9 months after diagnosis (19 ). Surgery was determined from both SEER and Medicare claims (for breastconserving surgery, International Classification of Diseases, 9th Revision [ICD-9], Procedure codes 85.20, 85.21, 85.22, 85.23, or and Current Procedural Terminology (CPT) codes 19110, 19120, 19125, 19160, or 19162; for mastectomy, ICD-9 Procedure codes 85.41, 85.42, 85.43, 85.44, 85.45, 85.46, 85.47, or and CPT codes 19180, 19182, 19200, 19220, or 19240) (20 24 ). The most extensive surgical procedure reported by SEER or Medicare during the initial treatment period was considered the definitive surgery. Method of axillary lymph node assessment was determined from SEER data. Treatment with radiation was determined from both SEER and Medicare claims (ICD-9 Procedure codes or 92.29; ICD-9 Diagnosis codes V58.0, V66.1, or V67.1; CPT codes or ; or Revenue Center Codes 0330 or 0333) (19,21 25 ). Patients were considered to have received radiation if either SEER or Medicare claims reported treatment with radiation during the initial treatment period. Treatment with adjuvant endocrine was not reported. Patient characteristics included age at diagnosis, race (white, black, white Hispanic, Asian Pacific Islander, or other unknown), year of diagnosis, marital status (widowed, married, single, separated divorced, or unknown), SEER registry, urban versus rural residence (large metropolitan, metropolitan, urban, small urban, or rural), median income of census tract or ZIP code ( 26 ), percentage of adults in census tract or ZIP code with less than 12 years education ( 26 ), and number of physician visits on separate days spanning an interval before diagnosis of 12 months to 1 month. A modified Charlson comorbidity index was calculated from Part A and Part B claims for an interval before diagnosis of 12 months to 1 month ( ). To enhance specificity, Part B diagnosis codes were included only if they appeared more than once over a period exceeding 30 days or in Part A claims as well ( 30, 31 ). Tumor characteristics included size, grade (low, intermediate, high undifferentiated, or unknown), histology (ductal not otherwise specified, comedo, papillary, cribriform, ductal with a lobular in situ component, or other) (32 ), and laterality (right or left). Tumor size as reported by SEER was determined from the pathology report, operative note, physical examination, mammogram, and ultrasound, with highest priority given to the pathology report and lowest priority given to the ultrasound. Margin status was not reported. The Hospital Cost Report Information Systems data were used to determine the teaching status of the hospital in which the initial surgery was performed (33 ). Statistical Analysis The risk of an event was estimated with the Kaplan Meier method. Patients were censored at the time of death, loss of feefor-service coverage, or loss of Medicare coverage. If none of these occurred, patients were censored at the end of A Cox proportional hazards model was used to determine the relationship between radiation and the risk of an event (defined as a subsequent ipsilateral in situ or invasive breast cancer reported by SEER and/or a subsequent mastectomy reported by Medicare claims) after adjusting for age, race, urban rural status, median income, marital status, comorbidity, size, histology, and grade (age and comorbidity were included due to clinical relevance; all other included variables had a univariate association with outcome at a P of.25 or less in unadjusted analysis). Age and size were treated as continuous variables on the basis of their linearity with outcomes. Missing values for all covariates were entered as dummy categories. The model was stratified by SEER registry and year of diagnosis to account for variations with geography and time. To evaluate the proportional hazards assumption, the logarithms of the hazard functions for the radiation and no radiation groups were compared by visual inspection. Prespecified interaction terms were tested to determine whether the effect of radiation varied with respect to age, size, histology, and grade. The analysis was repeated by use of a propensity score (i.e., the adjusted probability of receiving radiation ) to pair treated and untreated patients by use of a 1 : 1 matching algorithm (34 ). The results of the propensity score analysis were not substantively different from those of the Cox model, and therefore, only results from the Cox model are presented. A sensitivity analysis tested whether inclusion of patients with contralateral breast cancer or a change in the definition of the initial treatment period (from 9 months to 6 or 12 months) would alter effect sizes. In addition, we used the method proposed by Lash and Fink (35 ) to conduct a sensitivity analysis to determine how systematic error caused by the presence of an unmeasured confounder would alter the observed effect size of radiation. We created a binary variable to represent the presence or absence of the unmeasured confounder. The population prevalence of the unmeasured confounder assumed a uniform distribution. For each subject, the combination of exposure (treatment with radiation ) and outcome (a second breast cancer event) determined the probability of having the unmeasured confounder. A reconstructed dataset was created that included 1000 iterations of the original dataset, with each iteration having a population prevalence that was bounded by a prespecified uniform distribution. To determine how systematic error from an unmeasured confounder would alter the observed effect size of radiation, a Cox model adjusted for the unmeasured Journal of the National Cancer Institute, Vol. 98, No. 18, September 20, 2006 ARTICLES 1303

3 confounder and the original covariates was analyzed for each iteration of the reconstructed dataset. To determine how the combination of systematic and random error would alter the observed effect size of radiation, bootstrap estimates for the effect size of radiation were generated by resampling with replacement from the reconstructed data set. All statistical analyses were two-sided with an α of.05 or less and were conducted by use of SAS version 9.1 (Cary, NC). The Yale Human Investigations Committee approved this study and granted a waiver of informed consent. Patients were considered at high risk if they had any one of the following prespecified risk factors: age of years, tumor size of larger than 2.5 cm, comedo histology, and/or high grade (13 ). Patients without any of these risk factors were considered at low risk. The age cut point of 70 years was selected because current guidelines suggest that radiation may be omitted for women of age 70 years or older with early invasive breast cancer (11 ). A cut point of 2.5 cm was selected for size because a trial of patients with DCIS treated by lumpectomy that compares radiation with observation considered a tumor size of larger than 2.5 cm to be an exclusion criteria (36 ). The adjusted number needed to treat was calculated by dividing the unadjusted number needed to treat by the survival probability point estimate (37 ). The adjusted number needed to treat indicates the number of patients who require treatment with radiation to prevent one event after accounting for the competing risk of death. For example, if radiation conferred an 8% absolute improvement (unadjusted) in the risk of local relapse at 5 years and if 70% of patients were alive at 5 years, then the unadjusted number needed to treat would be 12.5 (i.e., 1/0.08) and the adjusted number needed to treat would be 18 (i.e., 12.5/0.70). R ESULTS Baseline Characteristics Of 3409 patients identified ( Table 1 ), median age was 74 years (interquartile range = years), 2818 (83%) were white, and 270 (8%) were black. Median tumor size was 0.7 cm (interquartile range = cm), 451 (13%) of the 3409 tumors were classified as high grade, and 640 (19%) were classified as comedo. Comorbidity was absent in 2303 (68%) of the 3409 patients, mild in 699 (21%), and moderate to severe in 294 (9%). A total of 1676 (49%) of the 3409 patients received radiation. Receipt of radiation was correlated with younger age, absence of comorbidity, comedo histology, and high grade ( Table 1 ). Concordance between SEER and Medicare claims was high for both radiation (92% concordance and κ = 0.84) and surgery (92% concordance and κ = 0.85). Unadjusted Risk of All Outcomes Median follow-up among the 3409 patients was 5.0 years (interquartile range = years). During follow-up, 63 (1.9%) patients experienced a second ipsilateral in situ breast cancer, 107 (3.1%) experienced a subsequent ipsilateral invasive breast cancer, 167 (4.9%) underwent subsequent mastectomy, and 498 (15%) underwent repeat breast-conserving surgery. was associated with a reduced risk for each outcome ( Table 2 and Fig. 1 ). The 5-year risk of a second breast cancer Table 1. Baseline characteristics * Covariate Total No., % event (the combined outcome) was 10.7% for women treated with conservative surgery alone versus 3.6% for women treated with conservative surgery plus radiation (difference = 7.1%, 95% confidence interval [CI] = 5.2% to 9.0%; P <.001). Consistent results were observed if the outcome were restricted to patients with either subsequent ipsilateral invasive breast cancer or subsequent mastectomy, with a 5-year risk of 8.6% for women treated with conservative surgery alone versus 3.3% for women treated with conservative surgery plus radiation (difference = 5.3%, 95% CI = 3.6% to 7.1%; P <.001). With regard to the secondary outcome of repeat breast-conserving surgery, the 5-year risk was 17.6% for women treated with conservative surgery alone versus 11.2% for women treated with conservative surgery plus radiation (difference = 6.4%, 95% CI = 3.8% to 9.0%; P <.001). Adjusted Effect of Therapy, % P Entire cohort Age, y < Race White Black Asian Pacific Islander White Hispanic Other or unknown Comorbidity index Unknown Tumor size, cm Unknown Histology DCIS, not otherwise <.001 specified Comedo Papillary Cribriform DCIS + LCIS Other Tumor grade Low <.001 Medium High undifferentiated Unknown * was defined as any code for delivery of radiation present in Medicare claims within 9 months of diagnosis or treatment with radiation reported by Surveillance, Epidemiology, and End Results data. DCIS = ductal carcinoma in situ; LCIS = lobular carcinoma in situ. P value is from two-sided Pearson s chi-square test. The logarithms of the hazard functions for the radiation and no radiation groups were parallel, indicating that the proportional hazards assumption was satisfied. was associated with a reduced risk of second ipsilateral in situ breast cancer (adjusted hazard ratio [HR] = 0.23, 95% CI = 0.12 to 0.45; P <.001), subsequent ipsilateral invasive breast 1304 ARTICLES Journal of the National Cancer Institute, Vol. 98, No. 18, September 20, 2006

4 Table 2. Five- and eight-year risks of any second breast cancer event and other outcomes * Follow-up Second in situ breast cancer Subsequent invasive breast cancer Subsequent mastectomy Any second breast cancer event 5 y 2.7 (1.9 to 3.5) 0.8 (0.3 to 1.3) 4.5 (3.4 to 5.5) 0.9 (0.4 to 1.3) 6.6 (5.3 to 7.9) 3.1 (2.2 to 4.1) 10.7 (9.1 to 12.3) 3.6 (2.6 to 4.5) 8 y 4.1 (2.8 to 5.4) 1.0 (0.4 to 1.7) 6.7 (5.2 to 8.2) 2.1 (1.2 to 3.1) 8.3 (6.6 to 9.9) 5.7 (4.0 to 7.4) 15.4 (13.0 to 17.8) 6.4 (4.7 to 8.2) * All outcomes are reported as the risk per 100 patients (95% confidence interval). Defined as a second ipsilateral, pathologically confirmed, in situ breast cancer reported by Surveillance, Epidemiology, and End Results (SEER). Defined as a subsequent ipsilateral, pathologically confirmed, invasive breast cancer reported by SEER. Defined as a subsequent mastectomy reported by Medicare claims. Defined as,, and/or. cancer (adjusted HR = 0.27, 95% CI = 0.16 to 0.45; P <.001), subsequent mastectomy (adjusted HR = 0.42, 95% CI = 0.29 to 0.60; P <.001), and any second breast cancer event (adjusted HR = 0.32, 95% CI = 0.24 to 0.44; P <.001) (Table 3 ). Other factors predicting an increased risk of a second breast cancer event included large tumor size, comedo histology, and high grade ( Table 3 ). Advanced age at diagnosis was associated with a lower risk of an event ( Table 3 ). The interaction of radiation with these factors was not statistically significant, indicating that the relative benefit associated with radiation was similar for patients at low risk and patients at high risk for an event. The risk reduction associated with radiation was not sensitive to inclusion of patients with contralateral breast cancer or to the definition of the initial treatment period. Inclusion of an unbalanced, unmeasured confounder reduced the effect size of radiation ( Table 4 ). An unmeasured confounder with the following characteristics negated the observed reduction in risk associated with radiation : 1) prevalence of 5% in the group receiving radiation and of 50% in the group receiving no radiation and 6.7-fold associated increase in event risk and 2) prevalence of 5% in the group receiving radiation and 85% in the group receiving no radiation and 3.7-fold associated increase in event risk. An unmeasured confounder that was more balanced between the two groups or was associated with a smaller effect size did not negate the statistical significance of the observed relationship between radiation and reduced event risk (Table 4 ). Absolute Effect of Therapy by Risk Group A total of 539 patients were low risk, 1570 patients were high risk, and 1300 patients could not be classified because of missing Fig. 1. Association of radiation (RT) with outcomes. A ) Risk of second in situ breast cancer, defined as a second ipsilateral pathologically confirmed, in situ breast cancer reported by Surveillance, Epidemiology, and End Results (SEER). B ) Risk of subsequent invasive breast cancer, defined as second ipsilateral pathologically confirmed, invasive breast cancer reported by SEER. C ) Risk of subsequent mastectomy, defined as a subsequent mastectomy reported by Medicare claims. D ) Risk of repeat breast-conserving surgery, defined as repeat breastconserving surgery reported by Medicare claims. The two-sided statistical test used was the longrank test. Vertical bars indicate 95% confidence intervals at 5 and 8 years. Journal of the National Cancer Institute, Vol. 98, No. 18, September 20, 2006 ARTICLES 1305

5 Table 3. Outcome predictors in adjusted analysis, hazard ratio (95% confidence interval) * Outcome predictor Second in situ breast cancer Subsequent invasive breast cancer Subsequent mastectomy Any second breast cancer event Breast radiation 0.23 (0.12 to 0.45) 0.27 (0.16 to 0.45) 0.42 (0.29 to 0.60) 0.32 (0.24 to 0.44) Age, y 0.94 (0.89 to 0.99) 1.0 (0.96 to 1.03) 0.96 (0.93 to 0.99) 0.97 (0.95 to 1.00) Race White 1.0 (referent) 1.0 (referent) 1.0 (referent) 1.0 (referent) Black 2.17 (0.87 to 5.43) 1.4 (0.64 to 3.23) 1.12 (0.59 to 2.13) 1.39 (0.85 to 2.29) Asian Pacific Islander 0.95 (0.31 to 2.91) 0.78 (0.24 to 2.53) 0.46 (0.18 to 1.22) White Hispanic 0.60 (0.076 to 4.71) 0.28 (0.038 to 2.11) 1.6 (0.76 to 3.55) 1.13 (0.56 to 2.28) Comorbidity score (referent) 1.0 (referent) 1.0 (referent) 1.0 (referent) (0.60 to 2.28) 1.4 (0.86 to 2.27) 1.13 (0.76 to 1.68) 1.2 (0.86 to 1.62) (0.20 to 2.30) 1.11 (0.51 to 2.40) 1.37 (0.72 to 2.23) 1.1 (0.70 to 1.80) Tumor size, cm 1.11 (0.85 to 1.46) 1.16 (0.98 to 1.38) 1.14 (1.01 to 1.29) 1.14 (1.02 to 1.26) Tumor histology DCIS, not otherwise specified 1.0 (referent) 1.0 (referent) 1.0 (referent) 1.0 (referent) Comedo 1.61 (0.79 to 3.26) 1.35 (0.80 to 2.26) 1.48 (0.99 to 2.22) 1.40 (1.00 to 1.97) Papillary2.00 (1.01 to 3.99) 1.4 (0.81 to 2.42) 1.09 (0.66 to 1.80) 1.41 (0.98 to 2.04) Cribriform 0.61 (0.078 to 4.76) 0.26 (0.04 to 1.90) 0.27 (0.06 to 1.11) DCIS + LCIS 1.21 (0.28 to 5.31) 1.24 (0.43 to 3.60) 1.28 (0.54 to 3.00) 1.39 (0.69 to 2.80) Grade Low 1.0 (referent) 1.0 (referent) 1.0 (referent) 1.0 (referent) Medium 1.47 (0.43 to 4.98) 2.12 (0.69 to 6.52) 1.68 (0.79 to 3.55) 1.49 (0.81 to 2.72) High 2.87 (0.81 to 10.26) 2.22 (0.65 to 7.57) 2.44 (1.11 to 5.38) 2.38 (1.24 to 4.56) * Model was adjusted for all patient, tumor, treatment, and hospital characteristics that were statistically significant in unadjusted analysis at α.25. The following variables are included in the model but were not reported because of limited clinical relevance and statistically nonsignificant P values: marital status, median income for patient census tract, and urban rural status. Model is stratified by Surveillance, Epidemiology, and End Results (SEER) geographic site and year of diagnosis. Entries in boldface indicate statistically significant results at P.05. DCIS = ductal carcinoma in situ; LCIS = lobular carcinoma in situ. Cells without an event have been left blank because the parameter estimate could not be calculated for these cells. De fined as a second ipsilateral, pathologically confirmed, in situ breast cancer reported by SEER. De fined as a second ipsilateral, pathologically confirmed, invasive breast cancer reported by SEER. Defined as a subsequent mastectomy reported by Medicare claims. Defined as,, and/or. was the referent group. data. Among low-risk patients, the 5-year event risk associated with conservative surgery alone was 8.2% and that associated with conservative surgery plus radiation was 1.0% (difference = 7.2%, 95% CI = 3.6 to 10.9; P <.001). Among high-risk patients, the 5-year event risk associated with conservative surgery alone was 13.6% and that associated with conservative surgery plus radiation was 3.8% (difference = 9.8%, 95% CI = 6.5 to 13.2; P <.001). Among unclassified patients, the 5- year event risk associated with conservative surgery alone was 9.3% and that associated with conservative surgery plus radiation was 4.1% (difference = 5.2%, 95% CI = 2.2 to 8.1; P <.001) (Table 5 and Fig. 2 ). Number Needed to Treat and Use of Therapy Within the high-risk group, patients of years with absent or mild comorbidity were most likely to benefit from radiation, with an adjusted number needed to treat of 11 patients to prevent one event in a 5-year period. For such patients, 65% actually received radiation. In contrast, Table 4. Sensitivity analysis to determine effect of an unmeasured confounder on the observed effect size of radiation * Prevalence of unmeasured confounder in patients who did not have the outcome, % (range) Unadjusted association between unmeasured confounder and outcome Effect size of radiation after accounting for unmeasured confounder, hazard ratio Systematic errorsystematic + random error Hazard ratio 2.50% 50% 97.50% 2.50% 50% 97.50% 5 (0 10) 70 (60 80) (0 10) 50 (40 60) (0 10) 50 (40 60) (0 10) 85 (80 90) (10 20) 70 (60 80) (10 20) 50 (40 60) (10 20) 50 (40 60) (10 20) 85 (80 90) * Outcome was a second breast cancer event, defined as a second ipsilateral, pathologically confirmed, in situ or invasive breast cancer reported by Surveillance, Epidemiology, and End Results and/or a subsequent mastectomy reported by Medicare claims ARTICLES Journal of the National Cancer Institute, Vol. 98, No. 18, September 20, 2006

6 Table 5. Five- and eight-year risks of any second breast cancer event for women at low and high risk for an event * No. of events per 100 persons (95% CI) No. P Absolute risk reduction Low-risk group y 1.0 (0.0 to 2.3) 8.2 (4.8 to 11.6) < (3.6 to 11) 8 y 1.0 (0.0 to 2.3) 14.8 (5.2 to 24.3) 13.8 (4.2 to 23.5) High-risk group y 3.8 (2.4 to 5.2) 13.6 (10.6 to 16.7) < (6.5 to 13.2) 8 y 7.8 (5.3 to 10.3) 20.4 (15.9 to 25.0) 12.7 (7.4 to 17.9) Unclassified group y 4.1 (2.2 to 6.0) 9.3 (7.1 to 11.5) < (2.2 to 8.1) 8 y 5.6 (2.9 to 8.4) 12.4 (9.6 to 15.3) 6.8 (2.8 to 10.8) Age years y 2.7 (1.2 to 4.2) 12.7 (8.8 to 16.6) < (5.8 to 14.2) 8 y 6.8 (3.7 to 9.9) 20.5 (14.2 to 26.8) 13.7 (6.6 to 20.7) Size >2.5 cm y 7.1 (0.2 to 14.0) 11.5 (1.9 to 21.1) ( 0.7 to 16.3) 8 y 11.3 (0.9 to 21.8) 16.7 (3.3 to 30.1) 5.4 ( 11.6 to 22.4) Comedo histology y 5.6 (3.1 to 8.1) 13.3 (8.5 to 18.2) (2.3 to 13.2) 8 y 11.4 (6.9 to 16.0) 19.0 (12.3 to 25.6) 7.5 ( 0.6 to 15.6) High grade y 4.5 (1.8 to 7.2) 20.3 (12.2 to 28.4) < (7.3 to 24.4) 8 y 5.5 (2.2 to 8.8) 22.2 (13.5 to 30.9) 16.7 (7.4 to 26.0) * Combined outcome of second ipsilateral in situ breast cancer, subsequent ipsilateral invasive breast cancer, and/or subsequent mastectomy. Number of women at risk at 5 and 8 years is reported in Fig. 2. CI = confidence interval. P value from two-sided log-rank test comparing radiation versus no radiation. All numbers are rounded to two significant digits. Therefore, the absolute risk reduction may not precisely equal the difference in the reported local relapse risks for radiation and no radiation. Low-risk group includes patients of age 70 years with tumor size of 2.5 cm, noncomedo histology, and non high grade. High-risk group includes any of the following characteristics: age years, tumor size >2.5 cm, comedo histology, and/or high grade. high-risk patients of 85 years or older with moderate to severe comorbidity were least likely to benefit from radiation, with an adjusted number needed to treat of 22. Only 27% of such patients received radiation ( Table 6 ). Patients in the low-risk group were less likely to benefit from radiation and to receive radiation than patients in the high-risk group. For low-risk patients of years with absent or mild comorbidity, the adjusted number needed to treat was patients treated with radiation to prevent one event in 5 years, and only 33% 54% of such patients received radiation. For low-risk patients of 85 years or older with moderate to severe comorbidity, the adjusted number needed to treat was 29, and none of these patients received radiation (Table 6 ). D ISCUSSION In this population-based study of older women with DCIS, radiation after conservative surgery was associated with a 68% reduction in the relative risk of a second breast cancer event. The relative benefit of radiation was highly consistent regardless of whether the endpoint was second in situ breast cancer, subsequent invasive breast cancer, or subsequent mastectomy and persisted despite the presence or absence of accepted risk factors such as younger age at diagnosis, large tumor size, high tumor grade, and aggressive tumor histology. The absolute benefit associated with radiation was considerable and compared favorably with that of other accepted clinical interventions. For example, 21 women with essential Fig. 2. Effect of radiation (RT) in low-risk and high-risk groups. An event was defined as any second breast cancer event, including a second ipsilateral, pathologically confirmed, in situ breast cancer reported by Surveillance, Epidemiology, and End Results (SEER) or a subsequent ipsilateral, pathologically confirmed, invasive breast cancer reported by SEER, and/or a subsequent mastectomy reported by Medicare claims. A ) Low-risk group defined as age of 70 years or older with tumor size of 2.5 cm, noncomedo histology, and low or intermediate grade. B ) High-risk group defined as having at least one of the following risk factors: age of years, tumor size of >2.5 cm, comedo histology, or high grade. The two-sided statistical test used was the long-rank test. Vertical bars indicate 95% confidence intervals at 5 and 8 years. Journal of the National Cancer Institute, Vol. 98, No. 18, September 20, 2006 ARTICLES 1307

7 Table 6. Adjusted number needed to treat to prevent one second breast cancer event * Age, y Charlson comorbidity index 5-Year overall survival, % (95% CI) Adjusted number needed to treat, (95% CI) % actually treated with radiation (No. receiving radiation/total No. diagnosed) Low risk High risk Low risk High risk (94 to 97) 11 (8 to 16) 65 (409/627) 1 93 (88 to 97) 11 (8 to 17) 59 (80/136) (66 to 89) 13 (10 to 20) 45 (22/49) (93 to 96) 15 (10 to 29) 11 (8 to 16) 54 (77/143) 70 (157/223) 1 88 (84 to 93) 16 (10 to 32) 11 (9 to 17) 52 (28/54) 71 (44/62) (58 to 79) 20 (13 to 41) 15 (11 to 22) 62 (13/21) 52 (11/21) (86 to 92) 15 (10 to 31) 11 (9 to 17) 47 (55/117) 63 (88/139) 1 90 (86 to 95) 15 (10 to 31) 11 (8 to 17) 33 (13/40) 58 (29/50) (57 to 78) 20 (14 to 41) 15 (11 to 23) (10/18) (77 to 86) 17 (11 to 34) 13 (9 to 19) 34 (23/68) 42 (33/79) 1 68 (60 to 77) 20 (13 to 41) 15 (11 to 22) 27 (6/22) 45 (13/29) (45 to 72) 24 (16 to 48) 17 (13 to 26) 0 (0/9) 35 (6/17) (56 to 74) 21 (14 to 43) 16 (12 to 24) (9/32) 1 66 (52 to 80) 21 (14 to 42) 15 (11 to 23) 0 (0/11) (26 to 69) 29 (19 to 59) 22 (16 to 32) 0 27 *CI = confidence interval. Low-risk group includes patients of age 70 years with tumor size of 2.5 cm, noncomedo histology, and non high grade. Cells for low-risk patients age years were left blank because all of these women were considered high-risk. High-risk group includes any of the following characteristics: age years, tumor size of >2.5 cm, comedo histology, and/or high grade. In concert with Surveillance, Epidemiology, and End Results Medicare guidelines, cell sizes less than five have been suppressed. Therefore, only the percentage of patients who received radiation is reported for these cells, not the raw numbers in the numerator and denominator. hypertension require 10 years of antihypertensive to prevent one cardiac event (38 ), and 36 women with osteoporosis require 3 years of bisphosphonate to prevent one clinically apparent fracture (39 ). Among healthy older women with DCIS, we found that 11 high-risk and low-risk patients required radiation to prevent one event at 5 years. These results strongly suggest that, for older women with DCIS, radiation should not be denied solely on the basis of chronologic age. Rather, the decision to recommend radiation should be based on two general considerations. First, an individual s risk for recurrence may be estimated, however coarsely, using established risk factors such as close or positive margins, high grade, the presence of comedo necrosis, and potentially other factors (9,13,15,40 45 ). Our study has confirmed that the absolute benefit associated with radiation was greater for high-risk patients than for low-risk patients. The second consideration is an individual s risk of intercurrent death before recurrence because a patient who dies before recurrence will not benefit from radiation. We have shown that, because of differences in the risk of intercurrent death, healthy women years old were twice as likely to benefit from radiation than women 85 years or older who have moderate to severe comorbidity. Given the growing number of elderly cancer patients (46 ), further research is urgently needed to evaluate the benefits of adjuvant cancer within the context of age, comorbidity, and risk of intercurrent death. During the study period, patients who were more likely to benefit from radiation were also more likely to receive radiation, suggesting that rational principles inform clinical decision making in the community setting. Nevertheless, among healthy older women at low risk for intercurrent death, 30% 40% of high-risk patients and 50% 70% of low-risk patients did not receive radiation. Further studies are needed to determine whether omission of radiation is driven by patient pref erence or by barriers such as physician bias against radiation or inadequate access to radiation facilities. Three randomized trials have demonstrated that, for women with DCIS, radiation after conservative surgery lowered the risk of recurrence by 38% 62% ( 4 6, 47 ). However, the vast majority of women enrolled in these trials were younger than 66 years, and outcomes for older women were not specifically reported. In contrast, a recent randomized trial conducted in older women with early invasive breast cancer treated with conservative surgery and tamoxifen concluded that radiation may not be needed for this patient group because the absolute risk reduction conferred by radiation was only 3% at 5 years ( 48 ). Similarly, recent observational data from Silverstein ( 9 ) suggested that radiation should be omitted for women 61 years or older with low-risk DCIS. In contrast, our data suggest that in the general population of older women with DCIS, ra diation after conservative surgery was associated with a more substantial clinical benefit, even for low-risk patients. Despite its observational nature, our study has several strengths. First, it represents the largest published series to evaluate the natural history of DCIS among women 66 years or older. Second, because this study was conducted with population-based data, its results are broadly applicable and demonstrate the external validity of prior randomized trials (49 52 ). Finally, our study has confirmed the prognostic significance of pathologic factors commonly available to clinicians and therefore highlights the importance of population-based data in the study of actual, if not ideal, clinical practice. Our study has certain limitations. For example, recurrence of in situ cancer within the ipsilateral breast may be underreported by SEER (17 ). In contrast, the more clinically significant endpoint of subsequent ipsilateral invasive breast cancer is more likely to be consistently reported by SEER because it represents a new histologic diagnosis. Given the potential for underreporting 1308 ARTICLES Journal of the National Cancer Institute, Vol. 98, No. 18, September 20, 2006

8 by SEER, we included Medicare-reported subsequent mastectomy as a marker to capture recurrence. This approach appears to have addressed possible problems with underreporting because the resulting risk of a second breast cancer event reported in our study was consistent with other large observational studies. For example, among patients treated with conservative surgery alone, our study reported a 5-year event risk of 10.7%, consistent with 5-year recurrence estimates ranging from 9% to 15.8% in previous studies (13,53 ). Among patients treated with conservative surgery plus radiation, our study reported a 5-year event risk of 3.6%, consistent with 5-year recurrence estimates ranging from 3% to 6% in previous studies (15,53 ). Another limitation concerns the absence of surgical margin status, a known risk factor for local recurrence (15,41,44 ). However, because patients with positive margins may be more likely to receive radiation, the absence of margin status would bias our results toward the null and result in the underestimation of the true benefit of radiation. Nevertheless, our results do not exclude the possibility that widely negative surgical margins may obviate the need for radiation in certain groups of older women, and this issue merits further study in prospective trials (10,44 ). In addition, although all patients in this study had pathologically confirmed DCIS according to SEER records, it is likely that, if each tumor was reviewed by a panel of expert breast pathologists, some tumors would be reclassified as benign proliferative lesions, lobular carcinoma in situ, or invasive cancer. Therefore, the effect size of radiation reported in this study may require further refinement when applied to patients diagnosed in centers with a dedicated breast pathologist. Finally, information on tamoxifen prescription and compliance was not available. Although tamoxifen has been shown to lower the risk of in-breast recurrence by approximately 3% at 5 years (54 ), this data were not published until 1999, and therefore, it is unlikely that many patients during the study period of January 1, 1992, through December 31, 1999, received tamoxifen. Even if all the patients treated with radiation also received tamoxifen, the small benefit of tamoxifen is insufficient to ex - plain the large benefit associated with radiation in this study (6,54 ). Our results indicate that, for older women with DCIS, radiation after conservative surgery was associated with a lower risk of second ipsilateral in situ breast cancer, subsequent ipsilateral invasive breast cancer, and subsequent mastectomy. Even for low-risk patients, the benefit of radiation compared favorably with other accepted interventions. Future studies should explore whether the apparent underutilization of radiation in this patient population reflects patient preference or stems from barriers to care. R EFERENCES (1) Ernster VL, Barclay J, Kerlikowske K, Wilkie H, Ballard-Barbash R. Mortality among women with ductal carcinoma in situ of the breast in the population-based Surveillance, Epidemiology and End Results program. Arch Intern Med 2000 ;160 : (2) Jemal A, Murray T, Ward E, Samuels A, Tiwari RC, Ghafoor A, et al. Cancer statistics, CA Cancer J Clin 2005 ;55 : (3) Berry DA, Cronin KA, Plevritis SK, Fryback DG, Clarke L, Zelen M, et al. Effect of screening and adjuvant on mortality from breast cancer. N Engl J Med 2005 ;353 : (4) Fisher B, Costantino J, Redmond C, Fisher E, Margolese R, Dimitrov N, et al. Lumpectomy compared with lumpectomy and radiation for the treatment of intraductal breast cancer. N Engl J Med 1993 ; 328 : (5) Julien JP, Bijker N, Fentiman IS, Peterse JL, Delledonne V, Rouanet P, et al. Radio in breast-conserving treatment for ductal carcinoma in situ: first results of the EORTC randomised phase III trial EORTC Breast Cancer Cooperative Group and EORTC Radio Group. Lancet 2000 ;355 : (6) Houghton J, George WD, Cuzick J, Duggan C, Fentiman IS, Spittle M. Radio and tamoxifen in women with completely excised ductal carcinoma in situ of the breast in the UK, Australia, and New Zealand: randomised controlled trial. Lancet 2003 ;362 : (7) Gotzsche PC. Re: Trends in the treatment of ductal carcinoma in situ of the breast. J Natl Cancer Inst 2004 ; 96 : 1257 ; author reply (8) Schwartz GF, Lagios MD, Silverstein MJ. Re: Trends in the treatment of ductal carcinoma in situ of the breast. J Natl Cancer Inst 2004 ;96 :1258 9; author reply (9) Silverstein MJ. An argument against routine use of radio for ductal carcinoma in situ. Oncology (Williston Park) 2003 ;17 : ; discussion , 1539, 1542 passim. (10) Wong JS, Gadd MA, Gelman R, Kaelin CM, Lester SC, Schnitt SJ, et al. Wide excision alone for ductal carcinoma in situ (DCIS) of the breast [abstract]. Breast Cancer Res Treat 2003 ;82 (Suppl 1):S10. (11) Carlson RW, Anderson BO, Burstein HJ, Cox CE, Edge SB, Farrar WB, et al. Breast cancer. J Natl Compr Canc Netw 2005 ;3 : (12) Morrow M, Strom EA, Bassett LW, Dershaw DD, Fowble B, Harris JR, et al. Standard for the management of ductal carcinoma in situ of the breast (DCIS). CA Cancer J Clin 2002 ;52 : (13) Kerlikowske K, Molinaro A, Cha I, Ljung BM, Ernster VL, Stewart K, et al. Characteristics associated with recurrence among women with ductal carcinoma in situ treated by lumpectomy. J Natl Cancer Inst 2003 ; 95 : (14) Solin LJ, Fourquet A, Vicini FA, Haffty B, Taylor M, McCormick B, et al. Mammographically detected ductal carcinoma in situ of the breast treated with breast-conserving surgery and definitive breast irradiation: long-term outcome and prognostic significance of patient age and margin status. Int J Radiat Oncol Biol Phys 2001 ;50 : (15) Solin LJ, Fourquet A, Vicini FA, Taylor M, Olivotto IA, Haffty B, et al. Long-term outcome after breast-conservation treatment with radiation for mammographically detected ductal carcinoma in situ of the breast. Cancer 2005 ;103 : (16) Vicini FA, Recht A. Age at diagnosis and outcome for women with ductal carcinoma-in-situ of the breast: a critical review of the literature. J Clin Oncol 2002 ;20 : (17) Warren JL, Klabunde CN, Schrag D, Bach PB, Riley GF. Overview of the SEER-Medicare data: content, research applications, and generalizability to the United States elderly population. Med Care 2002 ;40 :IV (18) Earle CC, Nattinger AB, Potosky AL, Lang K, Mallick R, Berger M, et al. Identifying cancer relapse using SEER-Medicare data. Med Care 2002 ; 40 :IV (19) Virnig BA, Warren JL, Cooper GS, Klabunde CN, Schussler N, Freeman J. Studying radiation using SEER-Medicare-linked data. Med Care 2002 ;40 :IV (20) Du X, Freeman JL, Warren JL, Nattinger AB, Zhang D, Goodwin JS. Accuracy and completeness of Medicare claims data for surgical treatment of breast cancer. Med Care 2000 ;38 : (21) ICD-9-CM: International Classification of Diseases, 9th Revision, 3rd Edition, Clinical Modification, Volumes 1, 2 and 3. Los Angeles (CA): Practice Management Information Corporation; (22) Buck CJ ICD-9-CM, volumes 1, 2, and 3 and HCPCS level II. Philadelphia (PA): W. B. Saunders Company; (23) Kirschner CG, Edwards NK, May DM, McNamara MR, Metge PJ, Pollack A, et al. Physicians Current Procedural Terminology. 4th ed. Chicago (IL): American Medical Association; (24) Anderson CA, Beebe M, Dalton JA, Duffy C, Evans D, Glenn RL, et al. Current Procedural Terminology CPT Chicago (IL): American Medical Association; (25) Du X, Freeman JL, Goodwin JS. Information on radiation treatment in patients with breast cancer: the advantages of the linked medicare and Journal of the National Cancer Institute, Vol. 98, No. 18, September 20, 2006 ARTICLES 1309

9 SEER data. Surveillance, Epidemiology and End Results. J Clin Epidemiol 1999 ;52 : (26) Bach PB, Guadagnoli E, Schrag D, Schussler N, Warren JL. Patient demographic and socioeconomic characteristics in the SEER-Medicare database applications and limitations. Med Care 2002 ;40 :IV (27) Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 1987 ;40 : (28) Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol 1992 ;45 : (29) Romano PS, Roos LL, Jollis JG. Adapting a clinical comorbidity index for use with ICD-9-CM administrative data: differing perspectives. J Clin Epidemiol 1993 ;46 :1075 9; discussion (30) Klabunde CN, Potosky AL, Legler JM, Warren JL. Development of a comorbidity index using physician claims data. J Clin Epidemiol 2000 ;53 : (31) Klabunde CN, Warren JL, Legler JM. Assessing comorbidity using claims data: an overview. Med Care 2002 ;40 :IV (32) Percy C, Van Holten V, Muir C. International Classification of Diseases for Oncology. 2nd ed. Geneva (Switzerland): World Health Organization; (33) Schrag D, Bach PB, Dahlman C, Warren JL. Identifying and measuring hospital characteristics using the SEER-Medicare data and other claimsbased sources. Med Care 2002 ;40 :IV (34) Parsons LS. Reducing bias in a propensity score matched-pair sample using greedy matching techniques. Proceedings of the Twenty-Sixth Annual SAS Users Group International Conference, Cary (NC): SAS Institute Inc; 2001 : (35) Lash TL, Fink AK. Semi-automated sensitivity analysis to assess systematic errors in observational data. Epidemiology 2003 ;14 : (36) Therapy and Oncology Group Phase III trial of observation +/ tamoxifen vs. RT +/ tamoxifen for good risk duct carcinoma in-situ (DCIS) of the female breast. Available at: pdf. [Last accessed: November 19, 2005.] (37) Walter LC, Covinsky KE. Cancer screening in elderly patients: a framework for individualized decision making. JAMA 2001 ;285 : (38) Wong ND, Thakral G, Franklin SS, L Italien GJ, Jacobs MJ, Whyte JL, et al. Preventing heart disease by controlling hypertension: impact of hypertensive subtype, stage, age, and sex. Am Heart J 2003 ;145 : (39) Black DM, Cummings SR, Karpf DB, Cauley JA, Thompson DE, Nevitt MC, et al. Randomised trial of effect of alendronate on risk of fracture in women with existing vertebral fractures. Fracture Intervention Trial Research Group. Lancet 1996 ;348 : (40) Rodrigues N, Carter D, Dillon D, Parisot N, Choi DH, Haffty BG. Correlation of clinical and pathologic features with outcome in patients with ductal carcinoma in situ of the breast treated with breast-conserving surgery and radio. Int J Radiat Oncol Biol Phys 2002 ; 54 : (41) Bijker N, Peterse JL, Duchateau L, Julien JP, Fentiman IS, Duval C, et al. Risk factors for recurrence and metastasis after breast-conserving for ductal carcinoma-in-situ: analysis of European Organization for Research and Treatment of Cancer Trial J Clin Oncol 2001 ; 19 : (42) Fisher ER, Dignam J, Tan-Chiu E, Costantino J, Fisher B, Paik S, et al. Pathologic findings from the National Surgical Adjuvant Breast Project (NSABP) eight-year update of Protocol B-17: intraductal carcinoma. Cancer 1999 ;86 : (43) Cutuli B, Cohen-Solal-le Nir C, de Lafontan B, Mignotte H, Fichet V, Fay R, et al. Breast-conserving for ductal carcinoma in situ of the breast: the French Cancer Centers experience. Int J Radiat Oncol Biol Phys 2002 ;53 : (44) Silverstein MJ, Lagios MD, Groshen S, Waisman JR, Lewinsky BS, Martino S, et al. The influence of margin width on local control of ductal carcinoma in situ of the breast. N Engl J Med 1999 ;340 : (45) Jhingran A, Kim JS, Buchholz TA, Katz A, Strom EA, Hunt KK, et al. Age as a predictor of outcome for women with DCIS treated with breastconserving surgery and radiation: The University of Texas M. D. Anderson Cancer Center experience. Int J Radiat Oncol Biol Phys 2002 ; 54 : (46) Federal Interagency Forum on Aging Related Statistics. Available at: [Last accessed: September 17, 2005.] (47) Fisher B, Land S, Mamounas E, Dignam J, Fisher ER, Wolmark N. Prevention of invasive breast cancer in women with ductal carcinoma in situ: an update of the national surgical adjuvant breast and bowel project experience. Semin Oncol 2001 ;28 : (48) Hughes KS, Schnaper LA, Berry D, Cirrincione C, McCormick B, Shank B, et al. Lumpectomy plus tamoxifen with or without irradiation in women 70 years of age or older with early breast cancer. N Engl J Med 2004 ; 351 : (49) Bijker N, Peterse JL, Fentiman IS, Julien JP, Hart AA, Avril A, et al. Effects of patient selection on the applicability of results from a randomised clinical trial (EORTC 10853) investigating breast-conserving for DCIS. Br J Cancer 2002 ;87 : (50) Murthy VH, Krumholz HM, Gross CP. Participation in cancer clinical trials: race-, sex-, and age-based disparities. JAMA 2004 ;291 : (51) Britton A, McKee M, Black N, McPherson K, Sanderson C, Bain C. Threats to applicability of randomised trials: exclusions and selective participation. J Health Serv Res Policy 1999 ;4 : (52) Gross CP, Mallory R, Heiat A, Krumholz HM. Reporting the recruitment process in clinical trials: who are these patients and how did they get there? Ann Intern Med 2002 ;137 :10 6. (53) Warren JL, Weaver DL, Bocklage T, Key CR, Platz CE, Cronin KA, et al. The frequency of ipsilateral second tumors after breast-conserving surgery for DCIS: a population based analysis. Cancer 2005 ;104 : (54) Fisher B, Dignam J, Wolmark N, Wickerham DL, Fisher ER, Mamounas E, et al. Tamoxifen in treatment of intraductal breast cancer: National Surgical Adjuvant Breast and Bowel Project B-24 randomised controlled trial. Lancet 1999 ;353 : N OTES Dr B. D. Smith is supported by the American Society of Clinical Oncology Young Investigator Award and the Breast Cancer Research Foundation. Dr C. P. Gross s efforts were supported by a Beeson Career Development Award (1 K08 AG24842) and the Claude D. Pepper Older Americans Independence Center at Yale (P30AG21342). Dr G. L. Smith is supported by National Institutes of Health/National Institute of General Medical Sciences Medical Scientist Training Grant GM The funding agencies had no role in the decision to publish or where to publish or the analysis. The authors assume full responsibility for these activities. This study used the linked SEER Medicare database. The authors are solely responsible for the interpretation and reporting of these data. The authors acknowledge the efforts of the Applied Research Program, NCI; the Office of Research, Development and Information, CMS; Information Management Services, Inc; and the SEER Program tumor registries in the creation of the SEER Medicare database. Manuscript received January 31, 2006 ; revised June 30, 2006 ; accepted July 20, ARTICLES Journal of the National Cancer Institute, Vol. 98, No. 18, September 20, 2006