Introduction

Bevacizumab is a humanized monoclonal antibody for vascular endothelial growth factor (VEGF), which is the most important regulator for angiogenesis in both healthy and pathological states [1]. Its enhanced expression is observed in many types of tumors and promotes tumor growth and metastasis [2]. Bevacizumab binds to VEGF, thereby inhibiting VEGF binding to VEGF receptors 1 and 2 on endothelial cells. The consequent inhibition of tumor angiogenesis at the tumor site is understood to suppress the growth of cancer cells [3]. Additionally, normalization of abnormal vessels in the tumor tissue reduces its interstitial pressure, thereby facilitating penetration by anticancer agents in combination with bevacizumab [4].

A meta-analysis on addition of bevacizumab to chemotherapy for patients with locally advanced or metastatic breast cancer (LA/mBC) showed that addition of bevacizumab to first- or second-line chemotherapy significantly prolongs progression-free survival (PFS) and overall response rate (ORR) but not overall survival (OS) [5]. However, another meta-analysis of factors indicating poor prognosis in patients with LA/mBC showed that addition of bevacizumab to first-line chemotherapy improves 1-year OS and OS in patients with poor prognostic factors, as compared with chemotherapy alone [6]. Regarding adverse events (AEs), addition of bevacizumab increases the incidence of hypertension, proteinuria, and bleeding; however, the incidence of thromboembolism or gastrointestinal perforation is unchanged and that of treatment-related deaths is low [5].

The JO19901 study, carried out in Japan, was a phase II study of bevacizumab plus paclitaxel in chemotherapy-naive patients with HER2-negative LA/mBC [7]. The primary efficacy endpoint, median PFS, was 12.9 months. Regarding secondary endpoints, ORR was 74% and median OS was 35.8 months. Regarding safety, no new serious AEs were detected. Thus, the study confirmed the reproducibility in Japanese patients of the efficacy and safety results achieved for bevacizumab plus paclitaxel combination therapy in studies conducted outside Japan.

Although several cohort studies have been carried out in other countries [8,9,10,11], clinical experience of bevacizumab plus paclitaxel combination therapy in Japan has been limited to the small number of patients in the JO19901 study, which enrolled 120 patients [7]. Therefore, we conducted a prospective multicenter observational study to investigate the effectiveness and safety of this combination as first- or second-line therapy for LA/mBC in daily clinical practice. Two cohorts, one comprising patients with hormone receptor-positive breast cancer and the other comprising those with triple-negative breast cancer, were established to enable comparison of prognostic factors in patients with each of these cancer subtypes and in patients receiving first- or second-line therapy.

Patients and methods

Study design

In this multicenter prospective observational cohort study, patients who met the following inclusion criteria were enrolled: histologically confirmed HER2-negative LA/mBC with confirmed HR status; Eastern Cooperative Oncology Group (ECOG) performance status (PS), 0–3; no history of second-line chemotherapy for LA/mBC; and sufficient bone marrow and major organ functions determined by the attending physician. Exclusion criteria included history of hypersensitivity to the ingredients of bevacizumab or paclitaxel, history of hemoptysis, uncontrolled hypertension, thromboembolism, positive urinary protein test result (≥ 2 +), gastrointestinal perforation, and severe fistula.

Patients were enrolled through central registration and classified by HR status: cohort A comprised patients with HR-positive breast cancer, and cohort B, those with triple-negative breast cancer (TNBC). First-line therapy was defined as treatment for patients who had not previously received chemotherapy for LA/mBC. Second-line therapy was defined as treatment for disease progression after or during receipt of first-line chemotherapy for LA/mBC. In cases of relapse during adjuvant chemotherapy, the first treatment after the relapse was considered the second-line therapy.

Written informed consent was obtained from all patients. The study protocol, procedures, and consent forms were approved by the institutional review board of each participating institution. The study has been registered with the University Hospital Medical Information Network Clinical Trials Registry (https://www.umin.ac.jp/ctr/index-j.htm; trial no. UMIN000009086).

Study treatment

Because the study was an observational study conducted in a clinical setting, dosage, treatment schedule, and criteria for dose reduction, interruption, and discontinuation were not specified. However, the study protocol recommended the following standard treatment regimen, which was used in the JO19901 study [7]: bevacizumab 10 mg/kg given every 2 weeks, and paclitaxel 90 mg/m2 given every week for 3 weeks, followed by a 1-week rest. Each combination of bevacizumab and paclitaxel administered as above for 4 weeks was deemed one cycle.

In cases of discontinuation of either drug due to AEs, the other drug could be continued as monotherapy. The protocol did not specify any treatment after discontinuation.

Study assessment

At screening on registration, medical history, symptoms of cancer, physical findings, pathological findings relating to the primary and metastatic lesions, presence or absence of measurable lesions, and previous treatments were recorded. During the treatment period, treatment schedule, treatment discontinuations, dose reductions, treatment interruption, concomitant drugs, and the last dosing date were recorded by electronic data capture.

Regarding safety, the incidence of five selected AEs of bevacizumab plus paclitaxel (i.e. neutropenia, hypertension, proteinuria, bleeding, and peripheral neuropathy), of any grade, was recorded. For other AEs, only those of grade ≥ 3 were recorded. AEs were evaluated based on CTCAE version 4.0 (Japanese Clinical Oncology Group edition) [12]. Effectiveness was evaluated and recorded in accordance with the Response Evaluation Criteria in Solid Tumors (RECIST), version 1.1 (Japanese Clinical Oncology Group edition) [13]. At the end of the observational period, patients’ disease progression, death, and post-treatment status were recorded.

Analysis populations and endpoints

The eligible patient population (used for the effectiveness analysis) was defined as patients who were registered according to the registration procedure, excluding those with ineligible cases or registration error. The treated patient population (used for the safety analysis and the sensitive analysis) was defined as patients who received bevacizumab plus paclitaxel combination therapy at least once. All evaluations were done by attending physicians.

The primary endpoint was OS, defined as the period between date of registration and death from any cause. Secondary endpoints were PFS, ORR, and safety. PFS was defined as the period between the registration date and the day when disease progression was determined (if that occurred first) or death (all causes).

Statistical analyses

This was an observational study conducted in the setting of daily clinical practice; therefore, the sample size was determined based on feasibility, considering the number of participating institutions, length of the registration period, and epidemiology of patients with HER2-negative LA/mBC. Consequently, the target numbers of patients were determined as 500 for cohort A and 250 for cohort B.

Expected median OS in each cohort according to treatment line (i.e. first- or second-line therapy) was estimated based on data from the prospective studies [7, 8, 14, 15, 17, 18]. Consequently, the expected median OS was 29.0 months and 18.0 months in patients receiving the study treatment as first-line and second-line therapy, respectively, in cohort A, and 17.0 months and 13.0 months in those receiving it as first-line and second-line therapy, respectively, in cohort B. Because the present study included patients who received the study treatment as both first- and second-line therapy, the ratio of first-line therapy patients to second-line therapy patients was assumed to be 5:5–7:3. Therefore, median OS was estimated to be 23.8 months in cohort A patients and 15.2 months in cohort B patients.

For the eligible patient population, cumulative survival curves for OS, median OS, and survival rate in each year were estimated using the Kaplan–Meier method, and Greenwood’s formula was used to construct 95% confidential intervals (CIs). Subgroup analysis was performed by Cox regression analysis to identify important prognostic factors. Sensitivity analysis was also performed, using data from the treated patient population. The same analyses were performed for PFS as those for OS. ORR was calculated as the proportion of patients achieving complete or partial response as the best overall response in patients with measurable lesions. CIs were calculated using the Clopper–Pearson method.

Safety was assessed using data from the treated patient population. The numbers of AEs, their grades, and their causal relation with the study drug were tabulated.

Results

Study population and baseline patient characteristics

A total of 767 patients were enrolled from 155 institutions across Japan between November 2012 and October 2014. Patient disposition is shown in Supplementary Fig. 1. Of these, the eligible patient population comprised 754 patients after exclusion of ineligible cases. Within this group, 539 (71.5%) were in cohort A and 215 (28.5%) in cohort B. The numbers of patients receiving the study treatment as first- and second-line therapy were 478 (63.4%) and 276 (36.6%), respectively. The treated patient population, that is, those who received the study treatment at least once, comprised 750 patients.

Baseline characteristics of the eligible patient population are shown in Table 1 and Supplementary Table 1a. Median age was 58 years. Most patients had distant metastasis (86.1%). Of these patients, most had visceral metastasis (91.7%), with ≥ 3 organs affected in a minority of cases (14.0%). Symptoms related to cancer (e.g. pain, dyspnea, pleural effusion, ascites, skin ulcer, and tumor fever) were experienced by 57.6% of eligible patients. Baseline characteristics of the treated patient population are shown in Supplementary Table 1c and are similar to those of the eligible patient population.

Table 1 Baseline characteristics (eligible patients)
Full size table

The proportions of cohort B patients with distant metastasis and metastasis to ≥ 3 organs were generally higher in those receiving first-line therapy than in those receiving second-line therapy; however, there were no differences for the other prognostic factors (Supplementary Table 1a).

Treatment exposure

Most patients received treatment in accordance with the treatment regimen used in the JO19901 study [7]. Details of treatment exposure in eligible patients are shown in Table 2 and Supplementary Table 2a. Median duration of bevacizumab and paclitaxel exposure was 5.1 and 4.9 months, respectively. Contrary to our expectation, duration of bevacizumab monotherapy after discontinuation of bevacizumab in combination with paclitaxel was extremely short and about 90% of cases discontinued bevacizumab at almost the same time as paclitaxel was discontinued (Table2, Supplementary Table 2a–c).

Table 2 Treatment exposure (eligible patients)
Full size table

Of the 754 eligible patients, 748 (99.2%) discontinued the study treatment; of these, 28.6% and 32.9% discontinued bevacizumab and paclitaxel, respectively, due to AEs. Regarding discontinuations due to other reasons, those recorded for ≥ 1% of patients included patient request (4.9%), maximum response (2.8%), breast surgery (2.7%), completion of scheduled treatment (2.3%), and treatment for other disease (1.5%).

The dose of bevacizumab or paclitaxel was reduced due to AEs in 1.5% and 33.1%, respectively, and it was suspended due to AEs in 19.7% and 29.7%, respectively.

When the treatment schedule of bevacizumab plus paclitaxel was the same as that in the JO19901 study [7], relative dose intensity of bevacizumab and paclitaxel was 99.2% and 90.9%, respectively.

Details of treatment exposure for patents in the treated patient population are shown in Supplementary Table 2b, c. Treatment exposure in this population was similar to that in the eligible patient population.

Effectiveness

Overall survival

Median observation period was 19.7 months. Events occurred in 496 of the 754 eligible patients (65.8%) during observation. Median OS was 21.7 months (95% CI 19.8–23.6 months), 25.2 months (95% CI 22.4–27.4 months), 13.2 months (95% CI 11.3–16.6 months), 24.4 months (95% CI 21.9–27.2 months), and 17.6 months (95% CI 15.2–20.0 months) in the full eligible patient population, in cohort A, in cohort B, in patients receiving the study treatment as first-line chemotherapy, and in those receiving it as second-line chemotherapy, respectively (Fig. 1a–c). Additionally, 1-year OS was 71.0%, 77.6%, 54.3%, 74.1%, 65.7%, in the full eligible patient population, in cohort A, in cohort B, in patients receiving the study treatment as first-line chemotherapy, and in those receiving it as second-line chemotherapy, respectively. Details of OS for eligible populations by cohort and treatment-line are shown in Supplementary Table 3 and Fig. 1d, e. Interestingly, OS was significantly longer in patients receiving the study treatment as first-line therapy than in those receiving it as second-line therapy in cohort A (log-rank test p < 0.0001, Fig. 2d), but not in cohort B (p = 0.3583, Fig. 1e).

Fig. 1
figure 1

Overall survival in the eligible patient population: a all eligible patients; b cohort A (patients with hormone receptor-positive breast cancer) versus cohort B (patients with triple-negative breast cancer); c, all eligible patients receiving first-line versus second-line therapy; d first-line versus second-line therapy in cohort A; e first-line versus second-line therapy in cohort B

Full size image

The results of multivariate analysis for OS in the eligible patient population are summarized in Table 3a. In decreasing order of hazard ratio (HR), the baseline characteristics independently associated with OS were TNBC, second-line therapy, ECOG PS ≥ 1, neoadjuvant or adjuvant taxane-based chemotherapy, cancer-related symptoms, and visceral metastasis.

Table 3 Results of univariate and multivariate analyses for overall survival
Full size table

The results of multivariate analysis in the 521 patients with recurrent breast cancer are summarized in Table 3b. The following baseline characteristics were identified as independent predictors of OS: ECOG PS ≥ 1, neoadjuvant or adjuvant taxane-based chemotherapy, visceral metastasis, cancer-related symptoms, and disease-free interval (DFI) ≤ 24 months.

The results of univariate and multivariate analyses of baseline characteristics associated with OS by cohort are shown in Supplementary Tables 4 and 5.

Progression-free survival and objective response rates

Median PFS was 8.5 months (95% CI 7.8–9.2 months), 9.4 months (95% CI 8.7–10.7 months), 6.0 months (95% CI 5.5–7.4 months), 9.3 months (95% CI 8.5–10.7 months), and 7.2 months (95% CI 6.0–8.4 months) in the full eligible patient population, in cohort A, in cohort B, in patients receiving the study treatment as first-line chemotherapy, and in those receiving it as second-line chemotherapy, respectively (Supplementary Fig. 2a, 2b, 2c). Details of PFS for eligible populations by cohort and treatment line are shown in Supplementary Table 6 and Supplementary Fig. 2d, 2e.

Multivariate analysis identified the following baseline characteristics independently associated with PFS (Table 4a): TNBC, ECOG PS ≥ 1, history of endocrine therapy, cancer-related symptoms, history of neoadjuvant or adjuvant chemotherapy, history of neoadjuvant or adjuvant taxane-based chemotherapy, and visceral metastasis.

Table 4 Results of univariate and multivariate analyses for progression-free survival
Full size table

Multivariate analysis also identified several baseline characteristics as independent predictors of prognosis in the 521 patients with recurrent breast cancer (Table 4b): ECOG PS ≥ 1, cancer-related symptoms, and DFI ≤ 24 months.

In the sensitivity analysis, the results for OS and PFS in treated patients were similar to those for the eligible patient population (Supplementary Tables 3 and 6).

ORR in patients with measurable lesions was 56.1%, 59.3%, 48.8%, 62.2%, and 45.1% in the full eligible patient population, in cohort A, in cohort B, in patients receiving the study treatment as first-line chemotherapy, and in those receiving it as second-line chemotherapy, respectively (Table 5a). ORRs by cohort and treatment line are summarized in Table 5b.

Table 5 Overall response rate in patients with measurable lesions
Full size table

Safety

The great majority of eligible patients (96.3%) experienced at least one AE, and 63.1% experienced one or more grade ≥ 3 AEs. Table 6 lists the AEs in treated patients. Incidences of grade ≥ 3 AEs hypertension, neutropenia, peripheral neuropathy, proteinuria, and bleeding were 35.7%, 27.2%, 7.2%, 3.7%, and 0.3%, respectively.

Table 6 Incidence of adverse events (AEs)
Full size table

Serious AEs were recorded in 66 patients (8.8%) including 15 patients with infection, five patients with congestive heart failure and 4 patients with drug-induced pneumonitis, fracture, gastrointestinal perforation, or liver dysfunction (Supplementary Table 7). Treatment-related deaths occurred in 6 patients (0.8%); the causes were liver failure (3 patients), acute gastroenteritis and heart failure (1 patient), gastrointestinal bleeding (1 patient), and gastrointestinal perforation (1 patient). Of the 3 deaths due to liver failure, one had liver failure associated with disease progression, and the other two had so-called pseudocirrhosis, which is associated with liver atrophy due to acute tumor response by chemotherapy on massive liver metastases and disorder of subsequent liver regeneration process.

Discussion

The B-SHARE study was a prospective observational study to investigate the effectiveness and safety of bevacizumab combined with paclitaxel as first- or second-line chemotherapy for HER2-negative LA/mBC under real-world clinical conditions in Japan. During the median observation period of 19.7 months, median OS for eligible patients was 21.7 months, and median OS for eligible patients receiving first-line therapy was 24.4 months. These results are within the range (21.6–30.2 months) achieved in previous phase III studies [14,15,16] and observational studies [8,9,10,11, 17]. Although there have been no previous observational studies on bevacizumab plus paclitaxel as second-line therapy, median OS for eligible patients receiving second-line therapy in the present study (17.6 months) was similar to the 18.0 months achieved in the RIBBON-2 study conducted as second-line chemotherapy [18], in which the efficacy and safety of bevacizumab combined with standard chemotherapy was compared with standard chemotherapy alone.

The 74.1% 1-year median OS for first-line therapy was a good result and similar to that determined by a meta-analysis of data from randomized controlled studies of bevacizumab combined with chemotherapy as first-line therapy (i.e. 71%) [6], showing that bevacizumab combined with chemotherapy may improve 1-year OS when compared with chemotherapy alone in high-risk patients.

The multivariate analysis results for OS in eligible patients identified TNBC, second-line therapy, poor PS, perioperative history of taxane therapy, cancer-related symptoms, DFI ≤ 2 years (i.e. recurrent breast cancer), and visceral metastasis as independent factors for poor prognosis. This is similar to the findings of previous studies on chemotherapy with [19] or without bevacizumab [20,21,22,23].

OS was significantly longer in patients receiving first-line therapy than in those receiving second-line therapy in cohort A but not in cohort B. Regarding baseline characteristics in cohort B, the proportions of patients with distant metastasis and metastasis to ≥ 3 organs were higher in those receiving first-line therapy than in those receiving second-line therapy, but no differences were found for the other factors. After completion of the study treatment, a greater proportion of patients receiving first-line therapy in cohort B were transferred to best supportive care compared with those in cohort A (33.1% and 21.2%, respectively). These findings suggest that patients with TNBC are less likely than those with hormone receptor-positive cancer to continue therapy because of many poor prognostic factors, but when patients were able to undergo second-line and subsequent therapy, they are likely to have a better prognosis.

As for first-line therapy, median PFS in eligible patients was 9.3 months and ORR in those with measurable lesions was 62.2%. As with OS, the results were consistent with those of previous randomized controlled studies [14, 15, 24, 25] and observational studies [8,9,10,11, 17]. For second-line therapy, median PFS was 7.2 months and ORR was 45.1%, similar to the results of the RIBBON-2 study [18].

The multivariate analysis results for PFS in eligible patients, including those with advanced disease, identified TNBC, poor PS, history of endocrine therapy, cancer-related symptoms, history of perioperative chemotherapy, history of perioperative taxane, and visceral metastasis as factors indicating poor prognosis. However, in patients with recurrent breast cancer, poor PS, cancer-related symptoms, and DFI ≤ 2 years were independent factors for poor prognosis. Therefore, poor prognostic factors for PFS differed with patient background. Although poor PS and cancer-related symptoms may be considered mutually associated, they were independent poor prognostic factors for both OS and PFS, regardless of whether the cancer was advanced or recurrent. The possibility that cancer-related symptoms are a poor prognostic factor in LA/mBC is supported by several other studies [23, 26, 27].

Despite the similarity in effectiveness (i.e. OS, PFS, and ORR) shown in the present study to that obtained in randomized controlled studies [14,15,16, 24] and observational studies [8,9,10,11, 17], the dosing period for bevacizumab in first-line therapy (5.3 months) was shorter than in randomized controlled studies [24, 25]. In fact, the bevacizumab dosing period tends to be shorter in observational studies [8,9,10,11, 17] than in randomized controlled studies [24, 25]. However, the bevacizumab dosing period in a retrospective cohort study [11] using information from the French Epidemiological Strategy and Medical Economics database was similar to that of the present study. The shorter dosing period in the present study compared with in randomized controlled studies may have been due to differences in patient selection (with poorer PS) and adherence to treatment. In the present study, 15.3% of patients were aged ≥ 70 years, and 8.0% had PS of ≥ 2. About 30% of patients discontinued treatment because of AEs, which is similar to that in the randomized controlled studies, whereas about 20% discontinued treatment without having disease progress (e.g. undergoing surgery after tumor shrinkage or switching to endocrine therapy).

No new AEs related to bevacizumab plus paclitaxel were detected in the present study. Incidence of all grades of AEs (96.3%) and those of grade ≥ 3 (63.1%) were higher than in previous randomized controlled studies [14, 18, 24, 25] and observational studies [8,9,10, 17]. However, there was no increase in the incidence of serious AEs or treatment-related deaths. We experienced 2 cases of treatment-related death due to so-called pseudocirrhosis during treatment of bevacizumab plus paclitaxel. Pseudocirrhosis is characterized by morphological changes in the liver that resembling cirrhosis on the radiological findings without typical histopathology of cirrhosis [28]. Pseudocirrhosis as adverse events by chemotherapy is not rare and an important complication of chemotherapy in patients with liver metastases. Recently, Oliai et al. [29] reported that pseudocirrhosis developed in 37 (55%) of 67 metastatic breast cancer patients with liver metastasis and was associated with poor prognosis in patients with live metastasis. They also described that chemotherapy agents associated with the development of pseudocirrhosis were albumin-bound paclitaxel, capecitabine, cisplatin, everolimus and vinorelbine. This adverse event is not bevacizumab-specific. However, the possibility that bevacizumab may inhibit the process of liver regeneration after treatment-induced hepatic injury cannot be ruled out.

The present study had several limitations. First, it was a single-arm observational study of bevacizumab plus paclitaxel combination therapy, so there was no direct comparison in terms of the effectiveness and safety between bevacizumab plus paclitaxel and paclitaxel alone. Second, treatment effectiveness (PFS and ORR) was assessed by attending physicians, and HR and HER2 status were also assessed at each facility. Central assessment or review was not done for the evaluation of effectiveness and those receptors status. Third, most patients received treatment in accordance with the treatment regimen used in the JO19901 study. Therefore, we could not examine the relationship between the dosage or the schedule of bevacizumab plus paclitaxel and its effectiveness to find the optimal use of this combination. Fourth, because the present study was done under daily clinical conditions, discontinuation due to the wishes of the patient or the decision of the attending physician was possible, regardless of whether the effects of treatment were sustained. During the course of treatment, various strategies were adopted after tumor reduction due to study treatment, such as discontinuation of treatment, switching to hormonal therapy for maintenance, or surgical intervention, which are uncommon in randomized controlled studies. The limitations of the present study make it difficult to obtain a true result for PFS and ORR. However, OS is a robust endpoint and we consider the OS reported here to be close to its true value, because it was achieved in patients treated with bevacizumab plus paclitaxel under real clinical conditions.

In conclusion, bevacizumab plus paclitaxel as first- or second-line chemotherapy in Japanese patients with HER2-negative LA/mBC was as effective as in previous randomized controlled studies and prospective observational studies. Furthermore, the good tolerability of this regimen was confirmed.