Inflammatory breast cancer (IBC) and patterns of recurrence

Understanding the biology of a unique disease




Inflammatory breast cancer (IBC) is the most aggressive manifestation of primary breast cancer. The authors compared the prognostic features of IBC and non-IBC locally advanced breast cancer (LABC) to gain insight into the biology of this disease entity.


This retrospective analysis consisted of 1071 patients, comprising 240 patients with IBC and 831 patients with non-IBC LABC who were enrolled in 10 consecutive clinical trials (5 from each disease group). All patients received similar multidisciplinary treatment. The authors measured time to disease recurrence for each individual site from the start of treatment to the date of disease recurrence or last follow-up (recurrence-free survival) and overall survival rates to the date of last follow-up or death.


The median follow-up period was 69 months (range, 1–367 months). Pathologically complete response rates were 13.9% and 11.7% in the IBC and non-IBC LABC groups, respectively (P = .42). The 5-year estimates of cumulative incidence of recurrence were 64.8 % and 43.4% (P < .0001), respectively, for IBC and non-IBC LABC. IBC had significantly higher cumulative incidence of locoregional recurrence and distant soft-tissue and bone disease. The 5-year overall survival (OS) rate was 40.5% for the IBC group (95% CI, 34.5%–47.4%) and 63.2% for the non-IBC LABC group (95% CI, 60.0%–66.6%; P < .0001).


IBC was associated with a worse prognosis and a distinctive pattern of early recurrence compared with LABC. These data suggested that investigating factors affecting “homing” of cancer cells may provide novel treatment strategies for IBC. Cancer 2007. © 2007 American Cancer Society.

Inflammatory breast cancer (IBC) is a rare but aggressive manifestation of primary breast carcinoma, with clinical and biological characteristics of rapidly proliferating disease.1, 2 The classic criteria required for clinical diagnosis, described by Haagensen,2 are diffuse erythema, edema involving more than two-thirds of the breast, peau d'orange (skin of an orange), warmth, tenderness, breast enlargement, and diffuse induration of the breast on palpation. IBC represents only 1% to 2% of primary breast cancer in the United States, but its incidence is dramatically increasing.3–5

Multidisciplinary management of IBC, consisting of induction chemotherapy, surgery, adjuvant chemotherapy, radiotherapy, and hormonal therapy (in hormone receptor-positive disease) has improved overall prognosis of this disease.6, 7 However, several retrospective studies that compared patients with primary IBC and those with non-IBC locally advanced breast cancer (LABC) have suggested that IBC is more aggressive clinically and biologically.7–9 These analyses are important, considering the rarity of IBC, because they suggest the need for specific guidelines in diagnoses and treatments. In fact, because of the rarity and the related difficulty in properly diagnosing IBC, patients with this entity are frequently treated with combined-modality regimens similar to those used in non-IBC LABC. Furthermore, despite differences in clinical and biological behavior associated with IBC and non-IBC LABC diagnoses, there are no currently approved IBC-specific treatments.

We hypothesized that because of such an aggressive clinical presentation, compared with non-IBC LABC, patients with IBC should have early establishment of occult micrometastatic disease. This phenomenon could be consequently associated with a distinctive pattern of recurrence that would explain the dismal prognosis associated with this diagnosis. This hypothesis necessitated an in-depth analysis of the pattern of recurrence of IBC compared with non-IBC LABC. Therefore, to test this hypothesis, we conducted a retrospective analysis that included a large population of patients with LABC who were treated at the University of Texas M. D. Anderson Cancer Center.10–17


Patient Population and Treatment Characteristics

A total of 1071 patients with either IBC or non-IBC LABC were treated on clinical trials at M. D. Anderson Cancer Center between January 1974 and August 2000. Patients were categorized into 2 groups on the basis of their clinical diagnosis of IBC or non-IBC LABC (Table 1). LABC was defined as stage IIB, IIIA, IIIB, or IIIC breast cancer by using 2002 American Joint Committee on Cancer (AJCC) staging guidelines.18 A clinical diagnosis of IBC required the presence of diffuse erythema, heat, ridging, or peau d'orange (corresponding to classification T4d in the AJCC system). The clinical diagnosis was confirmed for all patients by assessment of a multidisciplinary team comprising a medical oncologist, a surgical oncologist, a radiologist, and a radiation oncologist. Patients did not have evidence of metastatic disease on chest x-ray, abdominal CT, or bone scans. Diagnostic biopsy (eg, core or punch skin biopsy) were reviewed at the time of diagnosis by collaborating pathologists. Furthermore, all patients were treated in separate but parallel protocols with similar multidisciplinary approaches consisting of induction chemotherapy, locoregional treatment (surgery and radiotherapy), adjuvant chemotherapy, and hormonal therapy (for estrogen receptor-positive disease).10–12 The M. D. Anderson institutional review board approved the study and granted a waiver of informed consent because of the retrospective nature of the study. Clinical and pathology data were extracted from medical records.

Table 1. Baseline Patient and Disease Characteristics
 Locally advanced no. (%)Inflammatory no. (%)P
  • HR indicates hormone-receptor.

  • *

    Modified Black's grading system.

Total no.831 (100)240 (100) 
Age, y
 Median48 (—)51 (—) 
 [Range][15–79] (—)[27–78] (—)0.03
 Male2 (0.3)0 (0.0) 
 Pos354 (47.6)138 (57.5) 
 Pre388 (52.2)102 (42.5)0.02
T classification
 T011 (1.3)— (—) 
 T125 (3.0)— (—) 
 T2120 (14.5)— (—) 
 T3238 (28.8)— (—) 
 T4433 (52.4)240 (100)
N classification
 N0118 (14.3)38 (16.0) 
 N1269 (32.6)84 (35.3) 
 N2363 (44.0)89 (37.4) 
 N375 (9.1)27 (11.3)0.31
Clinical stage
 IIB88 (10.6)— (—) 
 IIIA253 (30.4)— (—) 
 IIIB490 (59.0)— (—) 
 INFLAM0 (0.0)240 (100)
HR status
 Negative249 (30.0)70 (30.0) 
 Positive350 (42.0)66 (27.0) 
 Unknown238 (28.0)104 (43.0)0.03
 119 (2.8)1 (0.6) 
 2214 (31.3)62 (36.9) 
 3451 (65.9)105 (62.5)0.12

In 5 consecutive multimodality clinical trials conducted between 1974 and 2000, 178 patients with IBC underwent anthracycline-based induction chemotherapy (3 clinical trials), and 62 patients were treated with anthracyclines and taxane-based regimens (2 trials).10–14 The first trial evaluated 5-fluorouracil, doxorubicin (Adriamycin), and cyclophosphamide (FAC) therapy administered as induction chemotherapy, followed by radiotherapy and further adjuvant chemotherapy with either FAC or cyclophosphamide, methotrexate, and 5-fluorouracil (CMF). The second trial used the same regimen of induction chemotherapy, followed by mastectomy, adjuvant FAC, and radiotherapy. Subsequently, in another trial, an adjuvant treatment consisting of vincristine and prednisone added to FAC was administered to patients who had a pathologically complete response (pCR) (defined as no evidence of residual invasive tumor in the breast and lymph nodes) after induction chemotherapy; this drug combination plus methotrexate and vinblastine was given to those who had a partial response, and methotrexate plus vinblastine alone was given to those who had less than a partial response. All patients received adjuvant radiotherapy to the chest wall and draining lymphatics and hormonal therapy if the tumor was hormone-receptor positive.

In 1994, paclitaxel was introduced for the management of IBC. Forty-four patients were treated with FAC as induction chemotherapy and adjuvant chemotherapy, and paclitaxel was given to patients with less than a partial response (only 16 patients) after FAC therapy and as adjuvant therapy in all patients.12 In a following pilot trial, 18 patients were treated by induction chemotherapy with sequential FAC, followed by 3 cycles of high-dose (175 mg/m2) paclitaxel weekly for 6 weeks, followed by 2 weeks of rest.13 In total, 240 patients were enrolled in 5 consecutive trials; 62 (26%) patients received paclitaxel, but only 34 (14%) received it as part of induction therapy. In the last 2 trials, patients with progressive or stable disease after induction therapy were offered high-dose chemotherapy with peripheral stem-cell support.14

The non-IBC LABC group comprised 831 patients enrolled in 5 consecutive clinical trials.15–17 The induction chemotherapy regimen in the first trial consisted of FAC combined with Bacille Calmette-Guerin (BCG); in the second trial, vincristine combined with doxorubicin (Adriamycin), cyclophosphamide, and prednisone (VACP) was used as induction chemotherapy. Subsequent trials used induction treatment with FAC followed by adjuvant FAC or methotrexate plus vinblastine based on the pCR to induction therapy; FAC (protocol IV) versus high-dose FAC (600 mg/m2 5-fluorouracil, 60 mg/m2 doxorubicin, and 1000 mg/m2 cyclophosphamide every 2 weeks with granulocyte colony-stimulating factor investigated the role of dose intensification; in the latest study, AT, doxorubicin, 50 mg/m2, and docetaxel, 75 mg/m2 introduced a taxane-based regimen as induction therapy for LABC. After surgery, all patients received adjuvant chemotherapy, radiotherapy, and hormonal therapy if indicated. In total, 831 patients were enrolled in 5 consecutive trials; 88 (10%) patients received a taxane as induction chemotherapy.

Statistical Analysis

Patient and disease characteristics were tabulated or described by their median and range. Characteristics were compared between the 2 disease groups with the chi-square test or Wilcoxon rank sum test, as appropriate. Median follow-up was calculated as both the median observation time among all patients and as the median observation time among patients still alive at their last follow-up. Recurrence-free survival (RFS) was measured from the start of treatment to the date of disease recurrence or last follow-up. Overall survival (OS) was measured from the start of treatment to the date of death from any cause or to the date of the most recent follow-up. Eight patients did not receive induction chemotherapy; thus, their survival time was calculated from the date of surgery. The survival times for the remaining patients were calculated from the start of induction chemotherapy. For the endpoint of RFS, patients who died before experiencing a disease recurrence were considered censored at their date of death. Survival distributions were estimated by the Kaplan-Meier product-limit method and the log-rank statistic was used to compare groups.

Multivariate Cox proportional hazards models were used to determine the association between disease group and OS and between disease group and PFS after statistical adjustment for patient and disease characteristics. Each model included terms for the disease group (IBC vs non-IBC LABC), hormone receptor status (positive vs negative), pathological response (pCR vs other), and number of positive lymph nodes. In addition, because patients were diagnosed over a wide period of time, year of diagnosis was included in the modeling. Hormone-receptor status was defined as estrogen and/or progesterone receptor-positive versus estrogen and progesterone receptor-negative disease. The presence of any residual disease in either the breast or lymph nodes after induction chemotherapy was considered less than a pCR. We estimated the cumulative incidence of any recurrence, soft tissue recurrence, bone recurrence, and visceral recurrence between the 2 disease groups by using the method of Gooley et al.19 In the analysis of any recurrence, death before disease recurrence was considered a competing risk. In the analyses of site of recurrence, recurrences at other sites, and death before recurrence were considered competing risks. For example, in the analysis of bone recurrence-free survival, patients who died before experiencing bone metastasis or experienced visceral, soft tissue, or unknown site of recurrence were considered to have a competing risk for bone recurrence. Note that 11 patients had recurrences at unknown sites. P-values were calculated according to the method of Pepe and Mori.20 The hazards of disease recurrence and death over time were estimated for each disease group by 12-month intervals and plotted by using a moving average smoother. We assumed a constant hazard after 5 years. P-values <.05 were considered statistically significant.


The median follow-up among all patients was 69.9 months (range, 1.3–367 months), and median follow-up among patients still alive at their last follow-up was 138.9 months (range, 18.3–366.8 months). Patients with non-IBC LABC included 490 (59%) patients with a diagnosis of inoperable clinical stage IIIB disease of which 433 (52.4%) patients had clinical T4 tumors. The percentage of patients with high nuclear grade was similar in both groups, whereas patients with non-IBC LABC were more likely to have hormone–receptor-positive disease, although in a significant number of cases the receptor status was unknown (28% among LABC and 43% among IBC patients; Table 1).

Induction chemotherapy consisted primarily of anthracycline-based regimens alone in 745 (89.7%) patients with non-IBC LABC and in 206 (86%) of IBC patients. The remaining patients received taxanes, in addition to anthracyclines, in the neoadjuvant or adjuvant setting (Table 2). There was no significant difference in clinically or pathologically complete response rates between the 2 groups. Patients with IBC who did not experience a pCR exhibited more frequently an advanced pathological stage (P < .0001). Furthermore, a higher number of involved lymph nodes was found in patients with IBC than in patients with non-IBC LABC (median number [range] of pathologically involved nodes, 4 [0–66] vs 2 [0–34], respectively; P < .0001).

Table 2. Post-treatment Patient and Disease Characteristics
 Locally advanced no. (%)Inflammatory no. (%)P
  • CR indicates complete response; Path CR, pathological complete response; XRT, radiotherapy; LN, lymph node(s).

  • *

    Patients enrolled in protocol A received neoadjuvant XRT and did not have surgery.

Systemic chemotherapy
Anthracycline-alone745 (89.8)206 (86.0) 
Anthracycline-taxane88 (10.2)34 (14.0).35
Clinical CR
 No742 (89.3)202 (87.1) 
 Yes91 (10.7)30 (12.9).31
Path CR
 No682 (88.3)155 (86.1) 
 Yes90 (11.7)25 (13.9).41
 Adjuvant XRT655 (82.6)188 (79.3) 
 None92 (11.6)35 (14.8) 
 Pre-op46 (5.8)14 (5.9).42
Surgery type
 Lumpectomy139 (16.8)0 (0.0) 
 Mastectomy634 (76.7)180 (79.3) 
 None*54 (6.5)47* (20.7)<.0001
LN removed
Positive LN

Four hundred twenty-nine patients (51.6%) with locally advanced disease and 163 (68%) patients with IBC experienced a recurrence (Table 3). Patients with IBC had a higher cumulative incidence of recurrence compared with non-IBC LABC (Fig. 1A). Furthermore, patients with IBC had a much higher cumulative incidence of bone and soft-tissue recurrence compared with patients with locally advanced disease (Figs. 1B, C, D). The latter was further analyzed and demonstrated a significantly higher incidence of both locoregional (skin, lymph nodes) and distant soft-tissue disease (Figs. 1E, F).

Figure 1.

Estimates of cumulative incidence of recurrence are shown by disease groups and site of recurrence (IBC, red line; LABC, blue line): (A) overall cumulative incidence of recurrence for IBC versus non-IBC locally advanced breast cancer (LABC); (B) cumulative incidence of bone recurrence; (C) cumulative incidence of soft tissue recurrence; (D) cumulative incidence of visceral recurrence; (E) cumulative incidence of local recurrence; and (F) distant soft-tissue recurrence.

Table 3. Estimates of the Cumulative Incidence of Recurrence
 No.No. eventsNo. competing events% 5-year estimate% 10-year estimateP
 Locally advanced8314297143.450.8 
 Locally advanced831824187.710.0 
Soft tissue
 Locally advanced831894119.410.6 
Local soft tissue
 Locally advanced831584426.67.0 
Distant soft tissue
 Locally advanced831294712.53.4 
 Locally advanced83125025026.129.7 

The 5-year RFS was 35.0% for all IBC patients (95% CI, 29.4%–41.8%) and 56.0% for non-IBC LABC (95% CI, 52.7%–59.5%) (P < .0001) (Fig. 2A). The 5-year OS was 40.5% (95% CI, 34.5%–47.4%) for IBC and 63.2% for non-IBC LABC (95% CI, 60.0%–66.6%) (Fig. 2B). Patients with IBC had decreased RFS compared with non-IBC LABC T4-classified patients (5-year PFS, 33.1%; 95% CI, 26.3%–41.6%, vs 44.7%, 95% CI, 40.2%–49.7%; P = .03) (Fig. 2C). Furthermore, patients with IBC had worse OS (5-year OS, 38.5%; 95% CI, 31.5%–47.2%) compared with non-IBC LABC T4-classified patients (5-year OS, 52.1%; 95% CI, 47.6%–57.0%) (P = .01) (Fig. 2D). Moreover, patients with IBC had worse 5-year RFS and OS compared with non-IBC LABC stage IIIB (stage IIIB, RPS, 53.5%; 95% CI, 48.9%–58.5%; P = .005; OS, 47.3%; 95% CI, 42.7, %–52.5%) (Fig. 2E, F). Subsequently, we excluded patients who received local radiotherapy as the only locoregional treatment, and we analyzed the 180 patients who were treated with surgery. One hundred seventeen patients experienced a recurrence, and the median RFS was 2.5 years (5-year RFS = 37.6%; 95% CI, 31.0%–45.6% and 10-year RFS = 34.0%; 95% CI, 27.5%–42.0%). One hundred fifteen of those died, and the median OS was 4.1 years (5-year OS = 45.9%; 95% CI, 39.0%–54.1%; 10-year OS = 35.7%; 95% CI, 28.9%–44.0%).

Figure 2.

Kaplan-Meier representation of relapse-free survival (RFS) rates in the 2 patients' groups are shown in (A) IBC (red line) versus LABC (blue line); (B) Kaplan-Meier representation of overall survival (OS) rates are shown in the same 2 patient groups. Kaplan-Meier representation of RFS is shown in 2 patients' groups: (C) IBC versus non-IBC LABC (T4 only); and (D) is the Kaplan-Meier representation of OS rates in the same groups. Kaplan-Meier representation of RFS in 2 patients' groups is shown for (E) IBC versus non-IBC LABC (stage IIIB) and (F) Kaplan-Meier representation of OS rates in the same groups.

Figure 3.

For IBC (red line) and LABC (blue line), plots of estimated hazard of (A) disease progression and (B) death are shown.

Plotting the estimated hazard of recurrence and death over time indicated that all patients with IBC had a higher risk for both events immediately after diagnosis, and then the hazard declined over time (Figs. 3A, B).

There was no statistically significant difference in RFS and OS survival between IBC and non-IBC LABC in the subset of patients that achieved pCR. However, patients with IBC that did not achieve a pCR demonstrated a 5-year RFS and a 5-year OS of 33.5% (95% CI, 26.7%–41.9%) and 40.9% (95% CI, 33.7%–49.6%), respectively, compared with 54.1% (95% CI, 50.5%–58%), and 63.1% (95% CI, 59.7%–66.8%) for patients with LABC (P < .0001).

Multivariate Cox proportional hazards for RFS and OS among all patients were performed (Tables 4, 5). Surgery type was evaluated for inclusion in the model, but no IBC patients received breast-conserving surgery, and in addition, surgery was collinear with the number of positive nodes and pCR. Similarly, pathologic stage was evaluated for inclusion, but it was collinear with pCR. After adjustment for age, HR status, positive axillary lymph nodes, use of induction chemotherapy, pCR, and year of diagnosis, patients with IBC had 1.60 times the risk of recurrence (P = .001) and 1.40 times risk of death (P = .02) compared with patients with locally advanced disease.

Table 4. Multivariate Models for Recurrence-Free Survival
 Hazard ratio95% Confidence intervalP
Inflammatory vs locally advanced1.601.22, 2.11.001
Age, continuous1.011, 1.02.24
HR positive vs HR negative0.780.63, 0.97.03
Path CR vs other response0.220.11, 0.44.00001
No. positive lymph nodes, continuous1.051.03, 1.06<.0001
Anthracycline + taxane vs anthracycline1.110.75, 1.65.61
Year of diagnosis, continuous0.960.93, 0.98.002
Table 5. Multivariate Models for Overall Survival
 Hazard ratio95% Confidence intervalP
  1. HR indicates hormone receptor; Path CR, pathological complete response.

Inflammatory vs locally advanced1.401.06, 1.86.02
Age, continuous1.011, 1.02.02
HR positive vs HR negative0.700.56, 0.87.001
Path CR vs. other response0.380.22, 0.65.0005
No. positive lymph nodes, continuous1.061.04, 1.07<.0001
Anthracycline + taxane vs anthracycline0.860.56, 1.31.48
Year of diagnosis, continuous0.960.94, 0.99.002


The current study represents the largest outcome analysis of patients with LABC. The major strengths of the study are the uniformly defined diagnosis of IBC and the homogeneity of systemic and local treatments. The analysis clearly demonstrated that IBC is a unique entity, characterized by more aggressive clinical features and a worse prognosis compared with non-IBC LABC. This observation remains valid even when comparing IBC only with LABC patients who have T4 disease, a clinical presentation that sometimes represents a challenging differential diagnosis with IBC, or LABC stage IIIB. Furthermore, patients with IBC who do not experience a pCR are more likely to have more extensive residual lymph nodal disease and a significantly worse prognosis compared with non-IBC LABC patients.21 Previous prognostic analyses for IBC have been quite difficult and have risen critical issues concerning appropriate criteria for diagnosis and the heterogeneity in treatments administered.3, 22 Furthermore, limited comparative analyses have not been able to define a pattern of recurrence that would explain the different prognosis for IBC.8

The higher risk of disease recurrence immediately after diagnosis and the distinctive pattern of soft-tissue relapse strongly support our hypothesis that these patients have already developed micrometastatic disease at the time of clinical diagnosis. These data are intriguing and suggest that unique considerations should be given to the development of novel diagnostic and therapeutic strategies for IBC. In fact, based on timing and pattern of recurrence demonstrated in IBC, we can speculate an initial “systemic phase” associated with early dissemination through breast lymphatics (probably a predominant mechanism) and blood vessels, with the establishment of occult microscopic disease (eg, bone marrow micrometastasis).24 The distribution or “homing” of microscopic disease appears to be not a random, but an orderly, predictable, and regulated multistep process responsible for the subsequent peculiar recurrence pattern seen in IBC. After a short clinical “dormant phase,” patients with IBC develop overt recurrent disease, and these events ultimately determine overall prognosis.

IBC has shown the capacity to spread early, primarily through lymphatic channels and secondarily through blood vessels.24–26 In fact, preclinical models have established that the higher expression of proangiogenic factors (eg, vascular endothelial growth factor) and a pattern of local tumor growth, called “vascular mimicry,” are characteristic features of IBC.25, 27 Increased expression of molecules involved in epithelial morphogenesis; regulation of intercellular adhesion (eg, E-cadherin), morphology, motility, and formation of cell junctions (Rho proteins, WISP3) are other molecular determinants of IBC.28, 29 Major biological factors that can contribute to understanding the early dissemination demonstrated for IBC and be responsible for the peculiar “homing” of cancer cells in local and distant sites are chemokines, molecules that are structurally and functionally similar to growth factors.30 Studies have shown that leukocyte chemoattractant receptors, named CXC chemokine receptor 4 (CXCR4) and CC chemokine receptor 7 (CCR7), are consistently expressed in breast cancer cells and associated with worse prognostic features and survival.30–32 We have recently demonstrated the expression of both chemokine receptors, CXCR4 (41%) and CCR7 (22.7%), in primary IBC.33 Moreover, there was a trend toward worse OS for patients with overexpression of cytoplasmic CXCR4; the prognostic value reached statistical significance when associated with epidermal growth factor receptor (EGFR) expression. These results suggest that the testing of therapies targeting lymphangiogenesis (eg, VEGFR-3 inhibitors) and CXCR4-inhibitors could produce novel therapeutic targets and provide more effective strategies in IBC.34, 35

The definitive demonstration of worse prognoses in patients with IBC suggests the need for a revision of the current staging system to better distinguish and separate this entity from non-IBC LABC.36 Furthermore, if it is accepted that patients with IBC have more diffuse disease at diagnosis, then more aggressive diagnostic and imaging strategies are needed to identify patients with more advanced disease, specifically diffuse microscopic disease.37, 38

In conclusion, our study demonstrates that IBC should be treated separately from non-IBC LABC and that the use of standard combinations of cytotoxic agents alone will not substantially modify the prognosis of patients with this disease. More sensitive diagnostic interventions and novel therapeutic strategies should be developed to increase the efficacy of systemic treatments in eradicating occult microscopic disease.