A Single-Center Experience With Inflammatory Breast Cancer, 1985-2003

Hypothesis  “Up-front” surgery improves survival in inflammatory breast cancer (IBC).

Design  Retrospective cohort, 1985-2003.

Setting  Tertiary referral center.

Patients  Consecutive patients with a primary occurrence of IBC.

Main Outcome Measures  All-cause and disease-free survival.

Results  One-hundred fifty-six patients were identified with IBC; 28 patients with metastatic disease were excluded from further analysis. The mean age of the remaining 128 patients was 53 years; 57% of women were postmenopausal. One hundred twenty-two patients had clinically apparent IBC. Tumors were palpable in 83 patients (mean diameter, 9.1 cm). Neoadjuvant chemotherapy was the initial therapy in 106 patients, while surgery was the initial therapy in 22 patients. The overall median survival was 37 months, with a median disease-free interval of 23 months. The 5-year survival was 42%, with a disease-free survival of 21%. Univariate analysis of recurrence identified previous hormone therapy (relative risk [RR], 0.50; P = .03), menopause (RR, 0.55; P = .01), and palpable adenopathy (RR, 1.57; P = .04) as significant factors. Univariate survival analysis highlighted previous hormone therapy (RR, 0.48; P = .04), radiotherapy (RR, 0.39; P = .02), sequence of therapy (P = .001), family history (RR, 0.47; P = .01), and palpable adenopathy (RR, 2.22; P<.001) as being important. Multivariate analysis of recurrence identified menopausal status as the key factor. Adenopathy at the initial examination was associated with decreased length of survival, while radiotherapy was associated with better survival.

Conclusions  Survival from IBC remains poor. Although adenopathy and radiotherapy affected survival by multivariate analysis, the sequence of therapy was not associated with improved outcome.

Inflammatory breast cancer (IBC) has an incidence of approximately 2.5 cases per 100 000 women¹ and is an especially aggressive form of locally advanced breast cancer. Inflammatory breast cancer is characterized by rapid tumor progression, early metastasis, and poor survival. Multimodal therapy (surgery, radiotherapy, endocrine therapy, and chemotherapy) has become the mainstay of treatment for IBC, offering a significant improvement in survival over single-modality or dual-modality treatments.² Multiple studies^1,3-9 have examined prognostic factors related to overall survival and disease-free survival in these patients. Previously identified prognostic factors include race, hormone receptor status, extent of erythema, menopausal status, tumor diameter, and lymph node involvement. In examining treatment-related factors, several series have suggested that response to induction chemotherapy may be an important prognostic factor.^2,4,6,10,11 Recent work has also suggested that the sequence of therapy (and more specifically, “up-front” surgery) may be a factor in survival.⁹ This has not been uniformly demonstrated,⁴ and the results of at least 1 study² suggest that surgery does not increase disease-free or overall survival when added to chemotherapy and radiotherapy protocols.

We sought to examine our single-institution experience with IBC during a 19-year period (1985-2003) to determine which prognostic factors might be identified for overall and disease-free survival. Special attention was paid to the sequence of therapy.

Following institutional review board approval, all patients seen at Mayo Clinic–Rochester for IBC between 1985 and 2003 were identified using our surgical and pathologic databases and the Mayo Clinic Tumor Registry. Inflammatory breast cancer was diagnosed on a clinical basis in patients who initially had erythema and edema of the breast with or without pathologic evidence of dermal lymphatic invasion. Preoperative biopsy specimens and operative specimens were evaluated for histologic evidence of dermal invasion. The site of preoperative biopsy (overlying the mass vs at the periphery) was unavailable. Patients without clinical signs of erythema and edema but with invasion of dermal lymphatics demonstrated on histologic examination were included. Patient records were evaluated if research authorization was available, if the patient had received treatment and follow-up at Mayo Clinic–Rochester, and if the patient was seen for a primary occurrence of IBC. Patient records were used to compile demographic, clinical, procedural, pathologic, treatment, and follow-up data. Patients were retrospectively staged based on the TNM criteria from the American Joint Committee on Cancer Cancer Staging Manual¹² (Table 1). Treatment sequence was determined by presentation, response to therapy, and ongoing research protocols. Early in the experience, neoadjuvant radiotherapy was more common based on a research protocol in place; later in the experience, neoadjuvant chemotherapy, followed by surgery, adjuvant chemotherapy, and radiotherapy became the treatment regimen of choice.

Recurrence and long-term survival were estimated using the Kaplan-Meier survival method for patients without evidence of metastases at the initial examination, and groups were compared using log-rank tests. All patients included in the recurrence and survival analysis had at least 3 years of potential follow-up. Time to recurrence and survival were calculated from the date of initial therapy. Cox proportional hazards regression models were used to select subsets of significant risk factors for each of the 2 outcomes.

Univariate analysis was performed using the χ² test for qualitative data and the Wilcoxon rank sum test for quantitative data to compare the patients who initially had surgery vs those who initially underwent treatment with chemotherapy. P<.05 was considered statistically significant for all analyses.

From 1985 to 2003, 156 patients (155 female) were identified as receiving treatment and follow-up at Mayo Clinic–Rochester for a primary occurrence of IBC. Twenty-eight patients (median survival, 18 months [range, 4-171 months]) had metastatic disease at the initial examination and were excluded from further analysis. For the remaining 128 patients, the mean age was 53 years (age range, 27-81 years), with most female patients having experienced at least 1 pregnancy (88%) or menopause (57%). Few patients had documented previous hormone therapy (20%) or a family history of breast cancer in a first-degree relative (22%). At the initial examination, 83 patients (65%) had a palpable mass, with a mean diameter of 9.1 cm (range, 3-20 cm) in greatest dimension. One hundred twenty-two patients (95%) had clinically apparent IBC. The remaining 6 patients had disease of a lower American Joint Commission on Cancer clinical stage but had pathologic evidence of IBC (invasion of dermal lymphatics). Clinical characteristics of the study group are summarized in Table 2.

Among 128 patients receiving chemotherapy, 106 (83%) had chemotherapy as the initial treatment. Overall, 122 study patients (95%) underwent surgery. Initial operations included modified radical mastectomy in 119 patients and wide local excision in 3 patients. Each patient who initially underwent wide local excision later underwent modified radical mastectomy. Of 22 patients having surgery as their initial treatment (followed by chemotherapy, radiotherapy, or both), operations included 19 modified radical mastectomies and 3 wide local excisions.

The dominant pathologic finding (88%) was infiltrating ductal adenocarcinoma. Pathologic results were unavailable for 2 patients: one did not have tissue confirmation, and the other had no residual disease at the time of resection and the preoperative biopsy results were from an outside institution and were unavailable. One hundred seventeen patients (91%) underwent radiotherapy, with a mean dose of 5500 rad (55 Gy) (median, 5900 rad [59 Gy]; range, 4100-7200 rad [41-72 Gy]). Radiotherapy was delivered with once-daily or twice-daily fractionation. Fields included were the affected chest wall, supraclavicular fossa, internal mammary lymph node chain, axilla with or without—a posterior axillary boost, and with or without—a boost to the surgical scar if delivered after surgery. Pathologic data are summarized in Table 3.

Among the 106 patients without distant metastasis who underwent neoadjuvant chemotherapy, regimens consisted of combinations of doxorubicin-cyclophosphamide in 53 patients (50%), doxorubicin-cyclophosphamide–5 fluorouracil in 20 patients (19%), cyclophosphamide–5 fluorouracil–prednisone in 13 patients (12%), doxorubicin-cyclophosphamide-methotrexate–5 fluorouracil in 13 patients (12%), cyclophosphamide-methotrexate–5 fluorouracil in 5 patients (5%), and doxorubicin-docetaxel in 1 patient (1%). One patient underwent endocrine therapy with letrozole.

Among 106 patients who underwent neoadjuvant chemotherapy, the cancers of 48 patients (45%) were successfully down staged. At the time of surgery, 16 such patients (15%) had a complete pathologic response of the primary tumor, including 4 patients with no histologic evidence of disease within axillary lymph nodes. Seventeen additional patients (16%) had only microscopic foci of primary disease at the time of surgery.

Among all 128 patients without evidence of distant metastases, the initial chemotherapy (neoadjuvant or adjuvant) was a doxorubicin-based regimen in 98 patients (77%). Compared with other regimens, doxorubicin-based therapy did not have a significant effect on recurrence (P = .93) or survival (P = .47).

Although 22 patients undergoing surgery as the initial therapy were of similar age and had similar evidence of palpable adenopathy compared with 106 patients undergoing chemotherapy as the initial therapy, there were differences (Table 4). For patients undergoing surgery as the initial therapy, the mean clinical tumor diameter was smaller (P = .02), and the clinical signs of dermal inflammation, specifically edema (P = .01) and erythema (P = .03), were less evident.

The sequence of therapy was examined as a risk factor. The sequences identified among 124 patients without metastatic disease at the initial examination and with at least 3 years of potential follow-up were of 3 types: trimodal in 102 patients (surgery-chemotherapy-radiotherapy in 13, surgery-radiotherapy-chemotherapy in 1, chemotherapy-surgery-radiotherapy in 60, and chemotherapy-radiotherapy-surgery in 28), bimodal in 19 patients (surgery-chemotherapy in 6, chemotherapy-surgery in 9, and chemotherapy-radiotherapy in 4), and unimodal in 3 patients (chemotherapy only). Although univariate analysis of time to recurrence showed no difference between sequence groups, a significant difference was seen with length of survival (P = .001) (Table 5). The survival curve is presented in the Figure.

The median survival for 128 patients with nonmetastatic disease was 37 months (range, 4-242 months). Eighty-five (66%) of 128 patients experienced a recurrence, with a median disease-free interval of 23 months. Recurrence was noted locoregionally (skin, surgical incision, chest wall, or axilla) in 27 patients. An additional 53 patients initially had systemic recurrence, including 5 patients with central nervous system metastasis. Among 22 patients who initially underwent surgery, 36% (8 patients) experienced locoregional recurrence, and 36% (8 patients) experienced systemic recurrence. In 106 patients who initially underwent chemotherapy, 19% experienced locoregional recurrence, 36% had systemic recurrence, and 7% experienced recurrence in the central nervous system (P = .46).

Five patients (4%) had progression of disease during treatment and never reached disease-free status. At the time of last follow-up, 51 patients (40%) were alive, while 77 (60%) had died. The 1-year survival was 92%. The 5-year survival for patients followed up for at least 5 years was 42%, while the disease-free survival was 21%.

Beginning in 1995, the treatment strategy of neoadjuvant doxorubicin and cyclophosphamide chemotherapy followed by mastectomy, adjuvant chemotherapy and radiotherapy, and endocrine therapy (in the case of positive receptors) became the most common treatment sequence at our institution. Before 1995, 60 study patients were treated; since 1995, 68 study patients were treated. Clinical and pathologic variables were similar between the 2 groups: the median tumor diameter (9 cm vs 8 cm), edema (83% vs 71%), erythema (83% vs 91%), pathologic N3 disease (28% vs 31%), and estrogen receptor (ER) positivity (52% vs 53%) were almost identical. The median survival for patients treated before 1995 was 39 months, with 1-year and 5-year survival of 95% and 40%, respectively. Among patients treated since 1995, the median survival was 35 months, with 1-year and 5-year survival of 90% and 45%, respectively. The 1-year and 5-year disease-free survival for the pre-1995 group were 75% and 17%, respectively, while the 1-year and 5-year disease-free survival for the patients treated since 1995 were 74% and 26%, respectively. Comparing the 2 eras (pre-1995 vs 1995 onward), there was no significant difference in 5-year overall or disease-free survival.

Univariate analysis was completed on the demographic, clinical, treatment, and pathologic variables, with the primary end points of time to recurrence and all-cause mortality among 124 patients without evidence of metastatic disease at initial examination and with at least 3 years of follow-up. Significant variables in time to recurrence were previous hormone therapy (relative risk [RR], 0.50; P = .03), menopause (RR, 0.55; P = .01), and palpable adenopathy at initial examination (RR, 1.57; P = .04). Variables identified as significant risk factors in time to mortality were previous hormone therapy (RR, 0.48; P = .04), radiotherapy (RR, 0.39; P = .02), sequence of therapy (P = .001), a positive family history (RR, 0.47; P = .01), and palpable adenopathy at initial examination (RR, 2.22; P<.001).

Results of the multivariate analysis were similar: menopause was associated with a decreased likelihood of recurrence (Table 6). Palpable adenopathy at initial examination was associated with decreased survival, while radiotherapy was associated with increased survival.

Inflammatory breast cancer remains a rare and aggressive disease process. Our retrospective single-institution experience with IBC identified 156 patients, 128 of whom were metastasis free at initial examination. Multivariate analysis suggests that one factor, menopausal status, was associated with a longer disease-free period (or a longer time to recurrence) among our patients with IBC. In the multivariate analysis of all-cause mortality, palpable lymphadenopathy at initial examination was the only factor associated with decreased survival, while radiotherapy was associated with increased survival.

In previous reports, several factors were associated with recurrence or survival. Variables previously identified as prognostic factors include race, ER status, menopause, extent of erythema, and lymph node metastasis.^1,3-7 Similarly, menopausal status and clinically apparent lymphadenopathy were prognostic factors in our study. Estrogen receptor status and progesterone receptor status were not independent predictors of recurrence or survival in our study, with approximately 50% of the patients with IBC at initial examination at our institution being ER positive (in agreement with other studies).^5,7,9,10 Previous hormone therapy was associated with decreases in recurrence and mortality in univariate analysis. The reason for this association is unclear, but among the subset of patients who had previously documented hormone therapy, a higher percentage had ER-positive tumors (60% vs 50%), making endocrine therapy a possibility (acknowledging that ER positivity itself was not a significant prognostic factor in our patient population). This factor, however, was not significant in the multivariate model, suggesting that it may simply identify those patients who are postmenopausal, which was a significant factor in our multivariate model of recurrence.

Recent attention has focused on the sequence of trimodal therapy in patients with nonmetastatic IBC, with varied results.^4,9,10 Liauw et al⁹ reported on 61 patients, 18 of whom had undergone up-front surgery compared with 43 who underwent neoadjuvant chemotherapy. They found that up-front surgery was associated with better cause-specific survival. This is in contrast to the large population-based study by Panades et al,⁴ in which 308 patients treated for curative intent were examined and were found to have no difference in relapse-free survival or cause-specific survival when comparing 35 patients who had undergone mastectomy as the initial therapy vs 148 patients in whom mastectomy followed neoadjuvant chemotherapy. Harris et al¹⁰ examined the sequencing of radiotherapy (before or after mastectomy) in a trimodal treatment regimen that included neoadjuvant chemotherapy and found no significant recurrence or survival difference. Similarly, previous work at our institution prospectively examined the outcome of preoperative radiotherapy (including accelerated fractionation) as part of a trimodal regimen in patients with nonmetastatic IBC.¹³ This study among 36 patients from 1983 to 1987 demonstrated a 5-year disease-free survival of 24%, with an overall survival of 34%.

In the present series, the sequence of therapy varied across the study period, with 22 patients receiving up-front surgery. Log-rank analysis of surgery as the initial therapy vs chemotherapy as the initial therapy did not demonstrate a significant difference, but when the patients were stratified by the specific sequence of therapy, this variable was significant in univariate analysis of mortality. Those patients initially undergoing surgery in a trimodal regimen, surgery-chemotherapy-radiotherapy, had a longer median survival (2560 days) than those undergoing regimens of chemotherapy-surgery-radiotherapy (2284 days) or chemotherapy-radiotherapy-surgery (954 days). In addition, those patients who received trimodal therapy (surgery, chemotherapy, and radiotherapy) fared better than those patients who received unimodal or bimodal therapy. These differences were not significant in the multivariate analysis and do not take into account any bias that may exist in the choice of sequence of therapy. For example, those patients who received preoperative radiotherapy after the initial treatment with chemotherapy were a mixed group. Early in this series, patients were treated on a protocol that involved neoadjuvant chemotherapy with cyclophosphamide–5 fluorouracil–prednisone and preoperative radiotherapy, regardless of their response to induction chemotherapy. Thirteen patients were treated with this protocol. Later in our experience, patients who underwent preoperative radiotherapy fell into 1 of 2 groups: (1) those patients who did not respond to neoadjuvant chemotherapy and (2) those patients who responded but had residual disease of significant size, causing concern about the ability to achieve clear margins at the time of mastectomy (both of which are potential sources of bias) and indicating less responsive or more aggressive tumors. In addition, specific and notable differences between those patients who initially underwent surgery vs those who initially underwent chemotherapy are summarized in Table 4. Although specific criteria for surgery as the initial therapy could not be identified from the medical records, these tumors tended to be smaller, with less obvious inflammatory signs (occult IBC). The patients receiving initial surgery were spaced throughout the study period and included several different staff surgeons, indicating that the decision to undergo up-front surgery was more likely related to the extent of disease at initial examination and not to individual staff preference.

Differences in clinical presentation are also apparent when comparing the subsets of patients undergoing the trimodal regimens of chemotherapy-surgery-radiotherapy and surgery-chemotherapy-radiotherapy. The mean tumor diameter at initial examination (9.4 cm for the group initially receiving chemotherapy vs 6.8 cm for the group initially receiving surgery) and the clinical signs of edema (75% vs 57%) and erythema (92% vs 64%) were different. There was no difference between groups in the percentage with adenopathy; however, taken as a whole, these differences complicate the interpretation of the survival analysis. Patients who initially received surgery in the sequence of therapy may simply identify those patients with a more favorable prognosis at initial examination and with less aggressive disease. In this series, we included patients with occult IBC, most of whom received up-front surgery. Although previous series have included patients with occult IBC, this may introduce bias.^9,14 Recent work has demonstrated better disease-free and overall survival for patients with intradermal tumor emboli and no signs of inflammation (occult IBC) compared with that of patients with intradermal tumor emboli and clinical signs of inflammation (clinical IBC),¹⁵ indicating that there may be a different biological process in those patients with clinical signs of inflammation.

The notion of different biological processes and unique subsets of patients with IBC is becoming more apparent as more molecular characterization is taking place. Several groups have examined genetic effects using various molecular techniques.^16-19 Recent work has confirmed the presence of distinct molecular subtypes of IBC,²⁰ and work from the same group has demonstrated differences in gene expression for patients with treatment-responsive tumors and those with tumors that were refractory to treatment.¹⁶ Additional work has examined the prognostic role of various genes, including p53 and MUC1.²¹

The treatment strategy described herein (neoadjuvant chemotherapy, mastectomy, adjuvant chemotherapy, and radiotherapy, followed by endocrine therapy if receptors are positive) has significantly increased overall survival compared with single-modality or dual-modality regimens,^2,22 but overall survival remains low compared with that associated with noninflammatory breast cancer. Overall survival in recent series ranges from 18% to 64%, depending on follow-up time and on specific treatment protocols.^2,4,5,7,9-11 In line with these previous experiences, our median survival was 37 months, with a 5-year survival of 42%. Comparing those patients treated before 1995, when updated treatment protocols were adopted, and those treated since 1995, little improvement was demonstrated in disease-free or overall survival, despite similarities in clinical presentation (nodal status, etc). Although the combined-modality regimen clearly provides the best outcome for patients with IBC, further research is necessary to delineate subsets of patients who may benefit from alterations in the approach to improve survival from this aggressive disease.

Correspondence: David R. Farley, MD, Department of Surgery, Mayo Clinic College of Medicine, 200 First St, SW, Rochester, MN 55905 (farley.david@mayo.edu).

Accepted for Publication: December 22, 2005.

Previous Presentation: This study was presented at the 113th Scientific Session of the Western Surgical Association; November 9, 2005; Rancho Mirage, Calif; and is published after peer review and revision. The discussions that follow this article are based on the originally submitted manuscript and not the revised manuscript.

Baiba J. Grube, MD, Galveston, Tex: Forty years ago, IBC was felt to be a nonsurgical disease because of prompt locoregional recurrence and because it was almost universally fatal. Radiotherapy as a sole modality was used to control regional disease. As chemotherapy regimens have achieved greater response rates in early-stage breast cancer, as well as locally advanced breast cancer, this experience has been applied to the treatment of inflammatory carcinoma. Surgery has returned to an integral part of management.

I am going to discuss 4 aspects of their study and ask 4 questions. In the presentation and in the manuscript, the authors proposed to test the hypothesis that the sequence of treatment has an impact on survival; that is, up-front surgery improves survival. The NSABP [National Surgical Adjuvant Breast Project] B-18 trial for stage I, II, and III noninflammatory breast cancer compared disease-free survival and overall survival in patients treated with chemotherapy prior to surgery compared to after surgery and demonstrated no difference in survival. My first question for the authors is, knowing these results, the sequencing of surgery has no impact on survival, and since this form of disease is almost universally systemic, what was the biological basis for the original hypothesis that up-front surgery would lead to improved survival?

In fact, the authors found no overall survival advantage to surgery before chemotherapy for inflammatory carcinoma. They did, however, identify a subgroup of patients with a significant difference in length of survival as a function of treatment sequence, with the up-front surgery cohort demonstrating prolonged survival by univariate analysis. Studies of inflammatory breast cancer are hampered not only by its rarity but by variations in the classification of the disease. Clinicians and pathologists have differed in the definition of inflammatory carcinoma. The distinction between local skin changes from an underlying locally advanced breast cancer or findings resulting from dermal lymphatic invasion from inflammatory carcinoma can be challenging even for experienced clinicians. The presence of dermal lymphatic invasion in the absence of clinical findings, an occult form of inflammatory carcinoma, has been recognized as a more favorable form of this disease. My second question, therefore, is was there a disproportionate diagnosis of inflammatory breast cancer made on the basis of dermal lymphatic invasion only in the absence of clinical findings in this subgroup of patients?

The authors found no improvement in survival with the anthracycline-based chemotherapy protocols initiated since 1995. However, the superiority of anthracycline-based chemotherapy compared to non-anthracycline regimens has been reported by the Early Breast Cancer Trialists' Cooperative Group meta-analysis for noninflammatory breast cancer. Moreover, recent results from clinical trials show that systemic therapy protocols that include taxanes, dose-dense chemotherapy, and (in some cases) biologics such as Herceptin for HER-2/neu–positive patients demonstrate additional improvement. My third question is what is your current systemic treatment protocol? What is the current sequencing protocol of these treatments? Have the results of this study changed the approach to this disease at your institution?

Finally, it has been found in the NSABP B-18 trial comparing preoperative chemotherapy to postoperative chemotherapy, as well as additional data from the NSABP B-27 trial, that the addition of taxanes to the anthracycline-based chemotherapy—and the large experience from M. D. Anderson with neoadjuvant chemotherapy for locally advanced and inflammatory breast carcinoma—that a complete pathologic response in the breast and in the axilla identifies a subset of patients with more favorable survival prognosis. My fourth question is have the authors separately determined the disease-free and overall survival in those patients that had a complete pathologic response in the breast and axilla and those who had a near-complete pathologic response?

Systemic therapy has returned to the surgical domain a disease that was once felt to be nonsurgical. Now that we are seeing complete pathologic responses in some patients, systemic therapy may again change our surgical approach.

Dr Farley: Your first question asked why we were concerned and interested in up-front surgery. Quite honestly, the paper from Florida (Liauw SL, Benda RK, Morris CG, Mendenhall NP. Inflammatory breast carcinoma: outcomes with trimodality therapy for nonmetastatic disease. Cancer. 2004;100:920-928.) was a stimulus to show us that perhaps there was some benefit to getting rid of the tumor early on. That was the real reason for doing the paper.

The second question involves occult IBC. Of the 22 patients who had surgery first, in fact 6 of them had a breast cancer that only was defined pathologically to be IBC. So finding occult IBC was uncommon, but it was certainly one of the factors for up-front surgery in this study. I think it is imperative to state that there is incredible bias with these patients. There are apples, oranges, pears, and bananas. This is a mixed group of patients that was difficult to treat, and sometimes the plan of therapy got changed in the middle of a woman's treatment regimen.

In regards to your third query, our Mayo protocol is to use adriamycin and Cytoxan (cyclophosphamide) up front, mastectomy second, Taxol (paclitaxel) therapy, followed by radiation therapy, and then based on HER-2/neu status and ER status whether Herceptin (trastuzumab) or tamoxifen is to be used.

It is difficult to assess whether, indeed, there was a complete response in each and every one of these 156 patients. We clearly show there were 16 women that had complete loss of any abnormality within the breast and no pathologic evidence of tumor. In fact, 4 of these 16 had no evidence of tumor spread within the lymph nodes. These complete responders were a small group, and we could not prove that a complete responder correlated with an obvious survival benefit.

Anees B. Chagpar, MD, Louisville, Ky: On univariate analysis, you stated that you found a significant difference in terms of survival based on the timing of surgery vs chemotherapy. But in fact, in your presentation, you showed a number of Kaplan-Meier curves accompanied by a single P value: surgery followed by chemotherapy followed by radiation, chemotherapy followed by surgery followed by radiation, chemotherapy followed by radiation followed by surgery, et cetera. Was the P value that you found on univariate analysis the comparison of all of those curves or was it simply the comparison of surgery followed by chemotherapy and radiation vs chemotherapy followed by surgery followed by radiation?

Dr Farley: There was no difference between patients having chemotherapy first and those having surgery first in the sequence of therapy. In the univariate analysis, you are exactly right: the 5 different lines that broke out were the 5 separate therapies that gave us a P value of less than .05.

Ravi Moonka, MD, Seattle, Wash: I hope this question is not a naive one, but what operation did these patients have? I assume it is some variation on a modified radical mastectomy, but in that operation the skin of the breast is used to close the incision, and in these patients the skin of the breast has cancer in it. My understanding was that is why chemotherapy is generally given initially, hoping to sterilize the margins of the tumor so you wouldn't be burdened by this high risk of local recurrence. Was it just a simple modified radical mastectomy or was a complex skin flap used to close the wound? Or did you simply not do surgery up front on patients in whom you were worried about a local recurrence?

Dr Farley: One hundred twenty-two patients underwent some form of surgery out of the 128. Three of those 122 had wide local excision elsewhere, came to the Mayo Clinic with positive margins, and they had modified radical mastectomy, as did 119 other patients having modified radical mastectomy. Within that group, a total of 8 patients had positive skin margins.