PATIENTS AND METHODS

The Breast Cancer Outcomes Unit database was initiated in 1989 to prospectively record diagnostic, prognostic, therapeutic, relapse, and survival data on all patients registered with breast cancer at the BCCA. The Breast Cancer Outcomes Unit database is maintained by the BCCA in Vancouver, Canada, and contains detailed demographic, staging, treatment, and outcome information for women referred to the BCCA since January 1, 1989. The database was used to select all women younger than 90 years of age who were referred to the BCCA between 1989 and 1995 with a diagnosis of EBC (tumor stage, T1 to T3; nodal stage, N0/1; metastatic stage, M0) and who had survived at least 30 days from the date of diagnosis.

The location of the breast primary was recorded as medial or lateral hemisphere. Because the study was designed to assess the prognostic significance of medial versus lateral location, patients with multicentric tumors, those for whom primary location could not be assigned to a medial or lateral hemisphere from the available information, and those with central, nipple, axillary tail, or areolar location, were excluded. Prognostic factors abstracted were age at diagnosis; tumor size (maximum histologic or gross pathology size in millimeters, or the clinical size from a preoperative mammogram or notes of the referring surgeon); lymphatic, vascular, or perineural space invasion (LVN) in the tumor (absent, present, unknown); pathologic axillary nodal status (pN0, pN1, unknown); estrogen receptor (ER) status (positive, negative, unknown), and tumor grade (nuclear grade or histologic grade using the modified Scarff-Bloom-Richardson system.¹⁶ Additionally abstracted were details of definitive surgery (breast-conserving surgery, mastectomy, or other), extent of RT, and systemic therapy delivered before diagnosis or suspicion of relapse.

The outcomes of interest were any relapse within the breast or chest wall (local recurrence [LR]), freedom from any inoperable regional or distant relapse or death from breast cancer (systemic disease-free survival [DFS]), and disease-specific survival. Patients for whom no follow-up details were available within 18 months of the analysis date (October 1998) were considered lost to follow-up and censored at the date of last contact.

RESULTS

Between 1989 and 1995, 7,697 patients were referred to the BCCA with tumor-node-metastasis stage I/II invasive breast cancer.²¹ Median follow-up was 3.9 years, and 11.6% (893 patients) were lost to follow-up. Actuarial local recurrence and systemic DFS at 5 years were 6.2% ± 1.5% (95% confidence interval) and 77.8% ± 0.8%, respectively. Disease-specific survival at 5 years was 84.9% ± 0.9%. Table 1 lists the distributions of prognostic factors and patient characteristics for the two cohorts. Of the entire study sample, 1,439 patients were excluded from further analysis because their tumors were central (583), in the axillary tail (101), or could not be definitively assigned to either the medial or lateral hemisphere (755).

Among patients who received RT, the majority received breast tangents or chest-wall fields only (85%); 424 (75 medial, 349 lateral) women were treated with a four-field technique (breast or chest wall, the axilla, and the supraclavicular nodes included in the therapeutic field but no intentional inclusion of the IMNs), and 94 (30 medial, 64 lateral) women received five-field RT (breast or chest wall, plus axillary and supraclavicular regions as above, and a direct field to include the IMNs). As listed in Table 2, medial lesions were smaller (mean diameter, 1.84 cm), less likely to have LVN invasion (LVN present in 33.0%), and less likely to be pN1 (25%) than lateral lesions (mean diameter, 1.98 cm; LVN positive in 38.1%; pN1 for 37%). Mean tumor grade was not statistically significantly different between medial and lateral locations.

Kaplan-Meier curves for systemic DFS and disease-specific survival are shown in Figs 1 and 2⇓, respectively. For both indices of outcome, the cohort that received systemic therapy had a higher event risk than the cohort of patients who did not receive systemic therapy. This is due to the poorer prognostic profile of the group of patients who received systemic therapy, as can be seen in Table 1. Among women who did not receive systemic therapy, no statistically significant univariate difference in systemic DFS or disease-specific survival was found for tumors with medial or lateral location. However, in the group of patients who received systemic therapy, systemic DFS and disease-specific survival were both lower for women with medial tumors.

The results of the Cox-regression multivariate analysis are listed in Tables 3 and 4⇓. For the no systemic therapy cohort, best-fitting prognostic models for systemic DFS and disease-specific survival included the same treatment/prognostic variables: local treatment, tumor size, LVN, tumor grade, and nodal status. For the systemic therapy cohort, the best-fitting prognostic model for systemic DFS included nodal status, tumor size, LVN, and tumor grade but not local treatment. For disease-specific survival, the best-fitting model for this cohort included only nodal status, tumor size, and tumor grade. In stage 2 of the Cox analyses, medial versus lateral location was added to the best-fitting prognostic models constructed in stage 1, yielding similar results to the univariate analyses shown in Figs 1 and 2⇑. In the no systemic therapy cohort, tumor location was not independently related to systemic DFS (P = .16) or disease-specific survival (P = .79). However, for women receiving systemic therapy, tumor location was statistically significantly related to both systemic DFS (P = .0001) and disease-specific survival (P = .002). In this latter cohort, the relative risks (multivariate) of systemic relapse and breast cancer death for medial as compared with lateral lesions were 1.53 ± 0.21 (95% confidence interval) and 1.46 ± 0.25, respectively.

To determine whether the observed difference in outcome between medial and lateral tumors in the systemic therapy cohort was related to the systemic therapy rather than the high-risk prognostic profile of the patients, modified St Gallen criteria⁴ were used to divide patients in each cohort into high-risk (pN1 or pN0 with at least one of tumor > 2 cm, LVN present, ER-negative) and non–high-risk (all others) subgroups. We used the St Gallen criteria because this system approximates better than does the tumor-node-metastasis staging system a population of patients for whom clinicians would today recommend adjuvant systemic therapy. Thus the findings would have greater relevance and could be more readily applied in the context of today’s adjuvant treatment strategies.

Univariate analyses of systemic DFS and disease-specific survival were conducted within each of the four resulting risk/treatment subgroups (high risk with and without systemic therapy; non–high risk with and without systemic therapy). Actuarial 5-year relapse rates for medial and lateral tumors within these four subgroups are listed in Table 5. Medial tumor location was not associated with a worse prognosis for three of the subgroups: women with low-risk tumors who did and who did not receive systemic therapy and women with high-risk tumors who did not receive systemic therapy. However, for women with high-risk tumors who did receive systemic therapy, medial tumor location was found to be a significant predictor of worse outcome compared with lateral location, with systemic disease-free survival rates of 66.3% and 74.2%, respectively (P = .003) and disease-specific survival rates of 75.7% and 80.8%, respectively (P = .028).

DISCUSSION

In this study, no significant difference was observed for local recurrence rates according to location of the primary tumor. Medial tumors were associated with a two-fold higher risk of systemic relapse and disease-specific death than lateral tumors in the group of women who received systemic therapy. No significant difference was observed for the group of women who did not receive systemic therapy. Furthermore, when analyzed according to tumor prognostic profile, the data for the systemic therapy group demonstrated a poorer prognosis for medial location only in tumors with adverse prognostic features. The observed relative risk of relapse (1.53) and death (1.46) for medial tumors is somewhat higher than that reported by Zucali et al.⁷ This is likely due to the differences in the analyses. Although both studies had a similar overall incidence of axillary node–negative disease (65.7% in the study of Zucali et al and 61.5% in this study), Zucali et al included central lesions, which are thought to have a worse prognosis, with medial lesions. Moreover, patients of low and high risk were analyzed together.

Information about tumor size, pN, grade, and LVN was missing in 5.5% to 7.8% of the cases in this series. These are relatively small numbers, given the overall sample size, and thus are unlikely to have influenced the outcome of the multivariate analysis. ER status was not available in 19.1% of cases (28.4% of the no systemic therapy cohort and 12.5% of the systemic therapy cohort). There is no biologically plausible explanation to support a difference in distribution of ER-positive and -negative tumors by location within the breast, and among the women for whom ER status was known, the proportion of ER-negative tumors was similar in both the medial and lateral hemispheres (P = .5 for entire cohort, P = .95 for the systemic therapy cohort, and P = .65 for the no systemic therapy cohort). Thus one might expect the distribution of ER-positive and -negative tumors to be similar among the women with medial and lateral tumors for whom this information was unknown. Their exclusion from the multivariate analysis might have weakened the significance of the observed results, given fewer numbers and therefore fewer events; however, it would not have artificially enhanced the difference in outcome between medial and lateral tumors, given the lack of association between location and ER status.

Studies have demonstrated that the combination of lumpectomy and RT results in similar rates of local relapse as mastectomy.^22,23 Local recurrence rates for all cohorts in this study were similar to those reported in these large randomized trials, which suggests that the patients treated by nonstandard methods (7% lumpectomy without RT, 7.2% treatment other than surgery) did not significantly alter the overall local recurrence rate for any cohort.

IMN recurrences are considered inoperable regional recurrences, and therefore they contributed to the systemic rather than local failure rate. Thus parasternal and IMN recurrences might be in part responsible for the lower DFS in high-risk medial lesions observed in this series. Several randomized studies^8,11 that compared survival and relapse for patients who received some or no local therapy to the IMN chain also failed to demonstrate a higher rate of regional or parasternal recurrence in the untreated IMN groups. However, it must be pointed out that recurrences in this region are generally apparent only with a computed tomography scan of the chest, a test not routinely used either at the time when these trials were conducted or in the metastatic work-up at the BCCA in the era that encompasses this patient cohort.

Patients who are considered to be at low risk of systemic relapse do not routinely receive adjuvant systemic therapy.⁴ Although the definition of low risk has changed over time, in general it has referred to small primaries, absence of axillary nodal spread, positive ER status, and low-grade histology. In addition, a small percentage of patients may not be offered systemic therapy because of comorbid illness, and some women with high-risk cancers may refuse recommended adjuvant therapy.²⁴ Consistent with these factors, the majority of women in the no systemic therapy cohort had low-risk features: 40% had tumors smaller than 2 cm, 86% had no LVN, and most significantly, 90% were pN0. This study was unable to demonstrate a difference in systemic DFS or overall survival rates for medial or lateral tumor location in this predominantly low-risk cohort with relatively few recurrences and deaths during the available follow-up period. Given the low event rate, a larger sample size and longer follow-up would be necessary to definitively demonstrate or exclude any difference in outcome that may exist.

It bears noting that in the no systemic therapy group, nodal status had a less significant relationship to outcome than did local treatment, tumor size, LVN invasion, and tumor grade (Tables 3 and 4⇑). This is misleading, given that nodal status is the most important prognostic factor in EBC.²⁵ However, in this cohort, 90% of the women had node-negative disease, and thus the influence of nodal status is based on events that occurred in the 10% of the cohort with positive nodes, whereas the significance of tumor size, LVN, and grade is determined on the event rate for the whole cohort.

Patients whose tumors have a favorable pathologic profile have a low risk of systemic micrometastases. In high-risk disease, however, micrometastases are a significant potential source of systemic relapse. This risk is reduced but not eliminated by the use of adjuvant systemic therapy.¹ In the absence of systemic therapy, disease from both IMN metastases and micrometastases elsewhere are eventual sources of systemic relapse, and the risk will be similar regardless of whether there has been spread to the IMNs before surgery. Thus even if there is a different frequency of IMN metastases among high-risk medial and lateral tumors, the disease-specific survival will be essentially the same. However, in the setting of systemic therapy, which reduces the relapse risk associated with micrometastases, differences in the frequency of IMN involvement for medial and laterally located tumors may result in different rates of systemic relapse and disease-specific survival, particularly when local control of the IMNs is not undertaken. This is a plausible explanation for the lack of difference in relapse observed for high-risk patients who did not receive systemic therapy regardless of tumor location, whereas high-risk patients with medial tumors who did receive systemic therapy had a worse outcome than their counterparts with lateral tumors.

Large randomized trials of radical mastectomy (RM) or extended RM in operable breast cancer (Table 6) have demonstrated higher rates of IMN metastases for pN1 medial lesions (32% to 34%) than pN1 lateral lesions (17% to 21.5%).^26-29 However, no difference was observed for medial and lateral locations among women with pN0 breast cancer. Additionally, retrospective reviews of large surgical series found worse 10-year overall survival rates for patients for whom both axillary and IMN nodes were positive (29% to 30%) compared with those for whom only axillary nodes were positive (55% to 60%) in the absence of adjuvant systemic therapy.^27,31 Mortality and distant metastases were reduced in axillary node–positive medial but not lateral lesions treated with surgical dissection and/or RT of the IMN basin in studies conducted in the pre–adjuvant therapy era. More recent prospective randomized trials in women with stage II/III disease who received adjuvant chemotherapy demonstrated a 30% reduction in the risk of death for the combination of modified RM and locoregional RT (inclusive of the IMNs) when compared with modified RM alone.^32,33 Unfortunately, there has been no subset analysis of outcome by primary site location in these more recent trials, and in at least one study, reported from our institution, the numbers do not permit such an analysis.

On examination of the RT given to the patients in this study, we found that the majority received RT to the breast or chest wall only, as part of breast-conserving surgery. A five-field technique was used more frequently than four-field RT in the earlier years of the study period. This change in practice arose from concern over late cardiac morbidity, and the increased complexity of matching a direct IMN field with the sloping medial tangent field in women who have had breast-conserving surgery.

The proportions of medial and lateral tumors treated with five-field RT were similar (2% and 1.7%, respectively). These patients were not excluded from the analysis for two reasons. First, the proportion of patients who received five-field RT was similar for medial and lateral lesions. Second, we reasoned that because IMN involvement is higher in high-risk medial than lateral lesions,^28-30 if local therapy (to the IMNs) reduces relapse and breast cancer–related death,^8,11 then inclusion of these patients would be expected to reduce rather than enhance the difference observed between medial and lateral tumors.

We found that patients who were treated with four-field radiotherapy had a worse outcome than those treated with five-field radiotherapy (data not shown); however, the numbers are too small to permit comparisons of outcome on the basis of RT technique in patients with medial-hemisphere cancers. In addition, the analysis is limited by its retrospective nature and the lack of prospective selection criteria used to determine the extent of RT. Nevertheless, combined with previously reported data, these findings support the hypothesis that optimal local control for high-risk medial lesions should include treatment of the IMNs.

Although surgical dissection in the hands of experienced surgeons is still the most reliable method to determine IMN status, it is unlikely that this approach will regain widespread enthusiasm. The use of this kind of surgery has declined with the increasing recognition of breast cancer as a systemic disease and the morbidity associated with complete dissection of the IMNs. Newer radiotherapy techniques that minimize cardiac and pulmonary morbidity are more likely to gain acceptance as an approach to local control of known or suspected IMN disease.^34,35

Whether RT of the IMN region will reduce the risk associated with high-risk medial breast cancers can only be definitively answered in a prospective fashion. There are ongoing randomized clinical trials (RCTs) comparing different RT techniques and field sizes after surgery in EBC. The European Organization for Research and Treatment of Cancer (EORTC) is testing the value of a direct field to the medial supraclavicular and ipsilateral IMNs (first three interspaces) in an RCT (EORTC 10925) with projected accrual of 4,500 subjects. The National Cancer Institute of Canada launched an RCT in 1999 comparing tangential fields alone with a modified four-field technique that includes the ipsilateral, upper IMNs in women with pN1 disease.

Our study, like that of Zucali et al,⁷ suggests that high-risk medial tumors are associated with significantly greater risk of relapse and death than high-risk lateral tumors. Although current practice does not include tumor location in decisions about extent of local therapy, it is hoped that long-term results from the ongoing RT trials will answer the question of the optimal volume of postoperative RT, particularly in high-risk, medial tumors, which seem to be associated with a higher risk of relapse and breast cancer death.