Fax: (720) 848-0222
Impact of postmastectomy radiotherapy in T3N0 invasive carcinoma of the breast
A Surveillance, Epidemiology, and End Results database analysis
Article first published online: 9 JUN 2008
Copyright © 2008 American Cancer Society
Volume 113, Issue 4, pages 683–689, 15 August 2008
How to Cite
McCammon, R., Finlayson, C., Schwer, A. and Rabinovitch, R. (2008), Impact of postmastectomy radiotherapy in T3N0 invasive carcinoma of the breast. Cancer, 113: 683–689. doi: 10.1002/cncr.23611
- Issue published online: 1 AUG 2008
- Article first published online: 9 JUN 2008
- Manuscript Accepted: 31 MAR 2008
- Manuscript Revised: 26 MAR 2008
- Manuscript Received: 14 FEB 2008
- breast cancer;
- postmastectomy radiotherapy;
Randomized trials provide evidence for improved outcomes with postmastectomy radiotherapy (PMRT) in high-risk patients. It has been suggested that patients with T3N0 breast cancer represent a favorable subgroup for which PMRT renders little benefit. In the current study, the authors used a United States population database to evaluate PMRT in this subgroup.
The cause-specific survival (CSS) and overall survival (OS) of women with T3N0M0 breast cancer in the Surveillance, Epidemiology, and End Results database after mastectomy and axillary staging from 1988 to 2002 were analyzed. Univariate analysis was performed to relate CSS with PMRT (yes vs no), tumor size (≤7 cm vs >7 cm), grade (1 vs 2 or 3), patient age (≤50 years vs >50 years), the number of lymph nodes dissected (≤13 vs >13), and the era treated (1988-1997 vs 1998-2002). Multivariate analyses for CSS and OS were also performed.
In total, 1865 women met the analysis criteria for OS; CSS data were available for 98.8% of those women. Of the women who were diagnosed during the era from 1988 to 1997, 22% received PMRT, and that rate increased to 41% during the era from 1998 to 2002. The actuarial 10-year CSS for those who received PMRT versus those who did not receive PMRT was 81.6% versus 79.8%, respectively (P = .38). PMRT was not associated with a CSS benefit in any subgroup, a finding that persisted in multivariate analyses. Women who received PMRT had an increased 10-year OS rate (70.7% vs 58.4%; P < .001) that was confined to women aged >50 years in a subgroup analysis.
This retrospective, population-based analysis demonstrated no increase in CSS with PMRT for women with T3N0 breast cancer, lending further support to the hypothesis that T3N0 disease postmastectomy represents a favorable subset of locally advanced breast cancer. Theincreased OS associated with PMRT in the absence of improved CSS likely reflects patient selection in this nonrandomized dataset. Prospective evaluation of PMRT in this population subset is warranted. Cancer 2008. © 2008 American Cancer Society.
The role of postmastectomy radiotherapy (PMRT) in the treatment of large (>5 cm), primary, lymph node (LN)-negative breast cancer remains incompletely defined.1, 2 Several randomized trials and meta-analyses have evaluated PMRT over the past several decades.3-11 However, those reports typically involved a very heterogeneous patient population, antiquated radiotherapy (RT) techniques and equipment, and suboptimal systemic therapy. Two prospective clinical trials garnered significant attention in 1997 after demonstrating not only a local control benefit but an overall survival (OS) benefit associated with PMRT in premenopausal patients with locally advanced disease who received systemic therapy.3, 4 A third trial demonstrated similar findings in postmenopausal patients shortly thereafter.5 Those trials had limited relevance to pathologic T3N0M0 (pT3N0M0) breast disease, as defined by the American Joint Commission on Cancer (AJCC) Cancer Staging Manual, 6th edition12 (stage IIB breast carcinoma, tumor >5 cm in greatest dimension without direct extension to the chest wall or skin [T3], no regional LN metastasis [N0], and no distant metastasis [M0]); the British Columbia trial excluded LN-negative patients entirely,3 and only a small minority of patients (8%-10% of the study population) were LN-negative in the 2 Danish trials.4, 5 Furthermore, it cannot be assumed that all LN-negative patients in the Danish trials had T3 disease, because pathologic involvement of pectoralis fascia and skin were features that allowed for enrollment, reflecting T2 and T4b tumor stages, respectively, by AJCC criteria.12 Detailed information confined to patients with pT3N0 disease was not provided in those publications. Moreover, the median number of axillary LNs evaluated in the trials that included patients with T3N0 breast cancer was less than that reported in most mastectomy series, suggesting possible under staging and subtherapeutic surgical management of the axilla.13 In view of the paucity of data regarding optimal postmastectomy therapy for T3N0 patients, we used the Surveillance, Epidemiology, and End Results (SEER) Database to assess the association between PMRT and survival within this TNM subgroup.
MATERIALS AND METHODS
The SEER Database
The SEER database is a national cancer registry that was established by the National Cancer Institute (NCI) and records patient, disease, and survival outcome (both overall and cause-specific) data for approximately 26% of the United States population. The SEER database currently is comprised of patient information from 18 geographically defined registries.14 The SEER registries routinely collect data on patient demographics, primary tumor site, extent of disease, first course of treatment, and follow-up. The first course of treatment before 1998 was defined in SEER as any treatment started within 4 months of the initiation of the first therapy. In 1998, it was changed to any therapy administered as a component of the planned treatment package (ie, surgery, chemotherapy, and PMRT) regardless of the duration of time since initiation of therapy. Data from these registries are deidentified and submitted electronically to the NCI on a biannual basis. Subsequently, the information is made available in the public-use SEER database. The SEER database does not include information relating to systemic therapy, surgical margins, or presence of lymphovascular and perineural invasion. Whereas SEER contains data regarding survival, locoregional recurrence data are not recorded.
The SEER database (SEER 17 Regs Limited-Use Nov 2006 Sub [1973-2004 varying]) was queried for patients who were diagnosed with pT3N0M0 ductal, lobular, or mixed breast cancer and underwent mastectomy and LN dissection or sentinel LN biopsy from 1988 to 2002. SEER*Stat software (version 6.3.5) was used to perform all queries. Patients coded as either receiving or not receiving PMRT were included; those without known radiation status were excluded. Patient age, tumor size and grade, and the number of LNs dissected were recorded for all identified patients. For the purpose of the analysis, grade 2 and 3 tumors were grouped together and compared with grade 1 tumors based on a published prognostic model that defined these 2 grades as high-risk for breast cancer mortality in the postmastectomy setting.15 Data points were retrieved from SEER using 6-month intervals for a maximum follow-up of 180 months.
Endpoints and Statistical Analysis
The primary endpoint of the analysis was cause-specific survival (CSS), which was ascertained by specifying breast cancer as the cause of death in SEER. CSS was measured from the time of diagnosis until death from breast cancer. Known nonbreast cancer deaths were censored and were not counted as events, and patients with unknown cause of death were excluded in the CSS analysis. OS was a secondary endpoint and was measured from the time of diagnosis until death from any cause.
Actuarial CSS and OS curves were generated by using the Kaplan-Meier method16 and were compared by using the log-rank test.17 The covariates analyzed included delivery of PMRT, age (<50 years vs ≥50 years), number of LNs dissected (less than or equal to the median number vs greater than the median number), tumor size (5.1-7 cm vs >7 cm), tumor grade (1 vs 2-3), and era of diagnosis (1988-1997 vs 1998-2002). Multivariate analysis using Cox proportional-hazards survival regression18 was performed to evaluate the influence of covariates on CSS and OS. Two-sided P values and 95% confidence intervals (95% CIs) are reported; statistical significance was defined as a P value ≤.05.
In total, 1844 women who met the defined inclusion criteria were identified for whom CSS data were available; OS data were available for 1865 women. Women with T3N0 breast disease who met the analysis criteria represented <0.3% of all breast cancer patients in the SEER database. Six percent of patients (120 of 1964 women) who otherwise would have met the inclusion criteria for CSS analysis were excluded because of unknown radiation status.
In total, 623 of the 1844 women (33.8%) analyzed received PMRT. The percentage of patients with T3N0 who received PMRT increased from 22% during the era from 1988 to 1997 to 41% during the era from 1998 to 2002. To evaluate for imbalances between the PMRT group and the no-RT group, patient, disease, and treatment factors are compared in Table 1. The PMRT group had significantly more patients aged <50 years and also had more patients with larger tumors (>7 cm).
|Variable||PMRT Group||No RT Group||P*|
|Tumor size, cm||.03|
|No. of LNs dissected†||.40|
Figure 1 shows that there was no statistically significant difference in CSS for patients who received PMRT versus those who did not (hazard ratio [HR], 0.88; 95% CI, 0.68-1.16; P = .38). Moreover, as shown in Table 2, there was no identifiable subgroup in which PMRT was associated with an increased CSS on univariate analysis. Univariate analysis of the association between each covariate pair and CSS is listed in Table 3. Several factors did predict for increased CSS, including age <50 years, grade 1 tumor, and diagnosis within the period from 1998 to 2002. In addition, there was a trend toward increased CSS for tumors that ranged in size from 5.1 cm to 7.0 versus tumors >7.0 cm (HR, 0.83; P = .17). Results of the multivariate analysis of factors that predicted for increased CSS are listed in Table 4. Age <50 years and grade 1 tumor remained statistically significant independent predictors of increased CSS on multivariate analysis, and there was a trend toward increased CSS associated with diagnosis during the era from 1998 to 2002 (P = .07). Like in the univariate analysis, delivery of PMRT was not associated with increased CSS in the multivariate analysis (HR, 0.94; 95% CI, 0.71-1.23; P = .64).
|Group||No. of Patients (%)||% 10-y CSS||HR*||95% CI||P†|
|All Patients||1844 (100)|
|No RT||1221 (66)||79.8|
|Age <50 y||536 (29)|
|No RT||303 (57)||83.5|
|Age ≥50 y||1308 (71)|
|No RT||918 (70)||78.5|
|Tumor size 5.1-7 cm||1261 (68)|
|No RT||855 (68)||80.8|
|Tumor size >7 cm||583 (32)|
|No RT||366 (63)||77.6|
|Grade 1||169 (9)|
|No RT||117 (69)||94|
|Grade 2-3||1675 (91)|
|No RT||1104 (66)||78.5|
|>13 LNs dissected||940 (51)|
|No RT||631 (67)||78.8|
|≤13 LNs dissected||904 (49)|
|No RT||590 (65)||81.2|
|Diagnosed 1998-2002||1104 (60)|
|No RT||649 (59)||‡|
|Diagnosed 1988-1997||740 (40)|
|No RT||572 (77)||79|
|Group||% 10-y CSS||HR*||95% CI||P†|
|Tumor size, cm|
|No. of LNs dissected|
|Tumor size, cm|
|No. of LNs dissected|
In contrast, PMRT was associated with increased OS. For the entire patient population, the actuarial 10-year OS rate was 70.7% for the PMRT group versus 58.4% for the no-RT group (HR, 0.65; 95% CI, 0.55-0.80; P < .001). However, the increase in OS was limited to patients aged ≥50 years, as shown in Table 5. The increased OS associated with PMRT was retained on multivariate analysis (Table 6). Age <50 years, grade 1 tumor, and greater than the median number of LNs dissected also were associated with increased OS. OS curves for patients in both age categories are shown in Figure 2.
|Group||% 10-y OS||HR*||95% CI||P†|
|All Patients (n=1865)|
|Age <50 y (n=540)|
|Age ≥50 y (n=1325)|
|Tumor size, cm|
|No. of LNs dissected|
To our knowledge, this is the largest analysis to date of PMRT in patients with T3N0 breast cancer. The rate of axillary LN involvement increases as breast cancer primary size increases,19, 20 with rates approaching 90% for T3 tumors according to a recent nomogram.21 Consequently, there are very scarce prospective data to guide oncologists toward optimal radiation management in the uncommon clinical scenario of LN-negative T3 disease. Updates of 3 large clinical trials published in the late 1990s supported the use of PMRT in select, high-risk patients.3-5 Two of those trials, conducted by the Danish Breast Cancer Cooperative Group4, 5 (trials 82b and 82c), included a small percentage of patients with LN-negative disease (267 of 3083 patients; approximately 9%). Both Danish trials demonstrated a 9% absolute improvement in OS with the addition of PMRT for patients with stage II/III disease who received systemic therapy. In subgroup analysis of the 82b trial (premenopausal women), the 10-year actuarial OS rate in LN-negative patients was improved to a statistically significant degree (82% vs 70%).4 In the 82c trial (postmenopausal women), in which 1 year of adjuvant tamoxifen was administered in both treatment arms, the difference in 10-year OS among LN-negative women was not significant (56% vs 55%).5 A common criticism of the 2 Danish trials is the relatively low median number of LNs identified and the high locoregional failure rates compared with historic standards for postmastectomy patients.22 Moreover, LN-negative patients with involvement of the pectoralis fascia (pT2) or skin (pT4b) were eligible for both Danish trials, and a subgroup analysis of only those patients with pT3N0 disease was not performed.
In 2001, an American Society of Clinical Oncology (ASCO) expert panel published guidelines for the application of PMRT, and it was ‘suggested’ that patients with T3N0 disease receive PMRT.23 Furthermore, guidelines published by the National Comprehensive Cancer Network (NCCN) recommend PMRT in patients with T3N0 disease.24 In practice, nearly 90% of radiation oncologists in North America and Europe recommend PMRT in the setting of T3N0 disease.1 Contrary to these guidelines and practice patterns, multiple retrospective series focusing on patients with T3N0 disease have suggested a high rate of locoregional control in the absence of PMRT, suggesting that large tumor size alone does not significantly increase the risk of locoregional recurrence.13, 25-27 Furthermore, a meta-analysis performed by the Early Breast Cancer Trialists' Cooperative Group demonstrated a small local control benefit of PMRT in node negative patients, which did not translate into a CSS benefit.28
The current analysis of patients registered in the SEER database supports the lack of a CSS benefit associated with PMRT in T3N0 breast cancer. Because locoregional failure after mastectomy has an ultimate impact on cancer-specific mortality,29 it would be very useful to analyze the relation between recurrence rates and the application of PMRT. Unfortunately, locoregional failure rates cannot be analyzed in this study population, because SEER does not contain recurrence data. The current study is limited in several other respects because of additional limitations in the SEER database. For instance, pathologic features that are associated with adverse outcome, such as lymphovascular or perineural invasion,19 are not available through SEER. Moreover, the SEER database does not contain information regarding surgical margin status or systemic therapy given in either the adjuvant or neoadjuvant setting. Hence, it is conceivable that an imbalance in prognostic factors or a disparity in the application of systemic therapy among the treatment groups masks a favorable impact of PMRT. Moreover, during the period from 1988 to 1997, PMRT was defined as a component of first course of therapy only if it was given within 4 months of surgery, thereby excluding those patients in whom PMRT was delayed beyond that length of time by protracted chemotherapy. However, in the context of this limited dataset, there is no suggestion of an association between PMRT and increased CSS for patients with T3N0 disease.
It is noteworthy that a statistically significant increase in OS limited to patients aged ≥50 years was associated with PMRT in the current analysis. Given the lack of a CSS benefit observed within the subgroup, the difference likely reflects selection bias in this nonrandomized patient population, with PMRT more likely to be delivered in postmenopausal patients who have a good performance status and fewer medical comorbidities. A recent SEER analysis of PMRT in elderly patients (aged ≥70 years) demonstrated an OS benefit associated with RT in ‘high-risk’ patients only, defined as those with T3/T4 primary tumors and/or N2/N3 disease.30 This has limited relevance to our analysis, because the T3N0 subgroup was not evaluated separately, and a CSS analysis was not performed.
Among the covariates that we examined, low tumor grade was associated with the most favorable outcome, with a 10-year absolute CSS increase of 13% relative to grade 2/3 tumors. It is noteworthy that larger tumor size (>7 cm vs 5.1-7 cm) was not a significant predictor of cancer-specific death on multivariate analysis. The era of diagnosis (1998-2002 vs 1988-1997) trended toward statistical significance, although inadequate follow-up in the more recent diagnostic era limits conclusions from this finding.
The use of PMRT for patients with T3N0 breast disease increased during the years analyzed. During the era from 1987 to 1997, approximately 22% of patients with T3N0 breast disease were treated with PMRT, and that rate increased to 41% during the era from 1998 to 2002. This likely reflects increased recognition of the putative benefit of PMRT in high-risk patients after updated publications of the Danish and British Columbia randomized trials in 1997 and 1998, although no specific data were presented in those publications with regard to the T3N0 population. It is interesting to note that <50% of women in the current analysis received PMRT even in the most recent ‘modern era.’ This low rate of PMRT administration is surprising in the context of ASCO and NCCN guidelines advocating its use in T3N0 patients23, 24 and a published patterns-of-care survey suggesting a much higher rate of PMRT use.1 This discrepancy may call into question the accuracy of the reporting of radiation delivery within SEER. However, a previous publication confirmed a high rate of accuracy of SEER radiation reporting, indicating that >94% of patients with Medicare claims for PMRT were coded correctly in SEER.31
In conclusion, our results from the SEER database lend further support for the position that women with T3N0 breast cancer who undergo mastectomy represent a favorable subpopulation of patients with locally advanced disease that is not associated with an increased CSS after the application of PMRT. However, conclusions derived from a population-based, nonrandomized dataset are limited, and the optimal postmastectomy treatment for women with T3N0 breast cancer would be clarified best through a prospective phase 3 clinical trial.
- 12GreeneFL,PageDL,FlemingID, et al, eds. American Joint Committee on Cancer Staging Manual,6th ed. New York, NY: Springer-Verlag; 2002.
- 14National Cancer Institute. Surveillance, Epidemiology, and End-Results Program. Available at: http://seer.cancer.gov. Accessed April 1, 2007.
- 18Regression models and life-tables. J R Stat Soc. 1972; 34: 187–202..
- 24National Comprehensive Cancer Network. NCCN Guidelines, version 2. 2007. Available at: http://www.nccn.org. Accessed September 24, 2007.
- 27Low locoregional recurrence rate among node-negative breast cancer patients with tumors 5 cm or larger treated by mastectomy, with or without adjuvant systemic therapy and without radiotherapy: results from 5 National Surgical Adjuvant Breast and Bowel Project randomized clinical trials. J Clin Oncol. 2006; 24: 3927–3932., , , et al.