Main

The development of resistance to standard endocrine therapy is a common problem for patients with oestrogen receptor-positive (ER+) metastatic breast cancer. Third-generation aromatase inhibitors (AIs) have shown superiority over tamoxifen in the treatment of both early and metastatic ER+ breast cancer (MBC) in postmenopausal women, and are now the most commonly used first-line endocrine therapy (Bonneterre et al, 2000; Nabholtz et al, 2000; Mouridsen et al, 2001; Milla-Santos et al, 2003; Paridaens et al, 2008). However, despite persistent expression of the ER in a majority of recurring tumours, even when an initial response is seen to standard endocrine therapies all tumours eventually develop resistance to endocrine therapy and the disease progresses (Johnston, 2010). Upon progression, patients are usually treated with second-line endocrine therapy or cytotoxic chemotherapy, which usually cause significant toxicities with response rates of only 10–20% (Chia et al, 2008; Di Leo et al, 2011).

Tumour response after withdrawal of endocrine therapy was first described following disease progression on tamoxifen treatment, and several subsequent case reports and small case series have reported similar responses to withdrawal of AIs (Legault-Poisson et al, 1979; Belani et al, 1989; Bhide and Rea, 2004; Cigler and Goss, 2006; Agrawal et al, 2011; Lambertini et al, 2011; Table 1).

Table 1 Retrospective experience with AI withdrawal
Full size table

Both in vitro and in vivo, breast cancer cells exposed to long-term oestrogen deprivation, comparable to that induced by aromatase inhibition in patients, become increasingly sensitive to growth stimulation by low levels of oestrogen and may have growth arrested in response to oestrogen concentrations that initially caused growth progression (Masamura et al, 1995; Shim et al, 2000; Song et al, 2001; Osipo et al, 2005). Thus, extremely low-oestrogen resultant upon ‘anti-oestrogen’ therapy such as AI therapy may sensitise the tumour cells to the low but normal levels of oestrogen found in postmenopausal women after AI withdrawal. Based on this hypothesis and the preclinical data related to extreme sensitivity to oestrogen after disease progression on AI therapy and because of the short half-lives of AIs (Guerrieri-Gonzaga et al, 2001; Buzdar et al, 2002), and supported by the abrupt resolution of AI-induced vasomotor symptoms, we conducted a prospective phase II clinical trial of observation after AI withdrawal for MBC.

Materials and Methods

Eligibility

Eligible patients were postmenopausal women aged ⩾35 years who had a diagnosis of histologically or cytologically confirmed hormone-receptor-positive (ER+ and/or progesterone-receptor-positive (PR+)) breast cancer with radiographically documented progressive or recurrent unresectable local-regional or metastatic disease. Prior treatment must have included AI monotherapy at the time of documented disease progression and this AI therapy was required to be discontinued at the time of study entry or <2 weeks prior to study entry. Patients who received AI therapy for MBC at the time of progression had to have had a prior radiographically documented response to AI therapy or to have been on therapy for at least 6 months. All patients were defined as postmenopausal at the time of enrolment by the following criteria: if >50 years, no menses within the past 12 months; or if <50 years, no menses within the past 12 months and FSH within postmenopausal range; or bilateral oophorectomies. All patients had to have been considered postmenopausal and eligible for prior AI therapy by their treating physician. Patients were eligible if they had measurable or non-measurable (but evaluable: defined as non-target lesions) disease according to modified Response Evaluation Criteria in Solid Tumors (RECIST 1.0; Therasse et al, 2000); patients with bone-only disease were also eligible; with Eastern Cooperative Oncology Group (ECOG) performance status of 0–2. Patients were excluded if they had life-threatening metastatic disease, had extensive hepatic involvement, had any past or present brain or leptomeningeal involvement, or symptomatic pulmonary lymphangitic spread, were highly symptomatic from their breast cancer; or if, in the opinion of their treating physician, they required urgent palliative chemotherapy. Prior or planned radiation therapy to a single site of evaluable disease in the event that the lesion was the only site of evaluable disease was also an exclusion criterion. Any number of prior endocrine or chemotherapy treatments for treatment of advanced breast cancer was allowed. However, concurrent alternative systemic anticancer therapy was not permitted. Concurrent bisphosphonate therapy was allowed if the therapy was initiated prior to enrolment.

Study design

The study was a single-arm phase II study of women who developed recurrent or progressive metastatic breast cancer while being treated with an AI. Within 2 weeks prior to study entry, all participants had to discontinue AI therapy and were subsequently actively followed off all therapy until disease progression or consent withdrawal. The primary end point was clinical benefit rate (CBR), which was defined as the proportion of patients with a complete response (CR), partial response (PR) or stable disease for a minimum of 24 weeks from the baseline point of AI withdrawal. The secondary end points were: progression-free survival (PFS) (defined as the interval between the date of AI discontinuation and the date of disease progression) and objective response rate (CR or PR) in patients who had measurable disease as defined by RECIST 1.0 at baseline.

The protocol was approved by the Dana-Farber/Harvard Cancer Center Institutional Review Board and written informed consent was obtained from all patients.

Study considerations

At the time of study entry, a complete medical history, physical examination, computed tomography scan of the chest, abdomen and pelvis, bone scans and blood work were obtained.

Once enrolled, participants were observed off all therapy until disease progression. Restaging studies including physical examination, blood chemistry, tumour marker (if available at participating institutions) and complete radiographic assessments were performed at 8, 16 and 24 weeks after AI discontinuation and every 12 weeks thereafter (Figure 1).

Figure 1
figure 1

Participants were observed off all therapy and evaluated for tumour response by examination, blood chemistry, tumour marker (if available at participating institutions) and complete radiographic assessment at 8, 16 and 24 weeks after AI discontinuation and every 12 weeks thereafter.

Full size image

Statistical analysis

The trial planned to enrol 40 patients in two stages in order to have a 90% power to detect a true CBR of 20% using a Simon’s two-stage design. In the first stage, 13 patients were to be accrued. If there was at least one patient with clinical benefit, an additional 27 patients were to be accrued. A 5% ‘lost to follow-up’ rate at 6 months was assumed.

The primary end point analysis consisted of the point estimate for the CBR and an exact binomial 95% confidence interval (CI) was reported. Patients lost to follow-up prior to 6 months, who were last known to be alive and progression free, were excluded from the CBR analysis. Estimates of PFS with their associated 95% CIs were calculated using the Kaplan–Meier product-limit method and Greenwood’s formula. All enrolled patients were included in the PFS analysis, regardless of length of follow-up.

Results

Patients

A total of 24 patients were enrolled from three centres (Massachusetts General Hospital, Beth Israel Deaconess Medical Center and Dana-Farber Cancer Institute) between October 2006 and December 2009, at which time the study was closed by the Dana-Farber Scientific Review Committee due to slow accrual. The baseline demographic and clinical characteristics of the patients are summarised in Tables 2 and 3. The median age was 63 years and the most common metastatic disease sites at baseline were bone in 21 (88%), lung in 9 (38%) and liver in 5 (21%) of the patients. Seven patients (29%) had bone-only disease. All patients received at least one metastatic or adjuvant prior endocrine therapy (metastatic setting 23 of 24 patients (96%), adjuvant setting 18 of 22 patients (82%)) with a median of 1 (range 1–4) prior endocrine therapies in the metastatic setting.

Table 2 Baseline patient demographics and clinical characteristics
Full size table
Table 3 Details of endocrine therapies prior to study enrolment
Full size table

Efficacy

At a median follow-up of 33.9 months, 22 subjects were evaluable for response. The remaining two patients had <24 weeks of follow-up, but were still progression free as of their date of last contact. The proportion of patients that achieved a CBR at 24 weeks was 46% (exact binomial 95% CI 26% to 67%), including 7 of 13 patients accrued for the first stage of the study. No CRs or PRs were observed. The median PFS of all patients was 4.47 months (95% CI 1.7 months to 17.2 months; Figure 2). Two patients have remained progression free over 60 months from the time of AI withdrawal on protocol, off all therapy. There were 14 reported deaths, occurred between 13 and 73.4 months following AI withdrawal, all of them related to disease progression. The median overall survival of all patients was 38.8 months (Figure 3).

Figure 2
figure 2

The median progression-free survival of all patients was 4 months (95% CI 2 months to 17 months).

Full size image
Figure 3
figure 3

The median overall survival of all patients was 39 months.

Full size image

Discussion

This first prospective trial designed to determine the withdrawal response rate after AI cessation demonstrated a CBR at 6 months of 46% (n=10 of 22). These results are encouraging and consistent with, and supported by, preclinical data in cell lines and xenograft models of oestrogen deprivation on disease progression (Masamura et al, 1995; Shim et al, 2000; Song et al, 2001; Osipo et al, 2005). The data show that cells deprived of oestrogen are able to adapt to anti-oestrogen therapy and eventually may become growth-stimulated rather than inhibited. In ER+ cell lines subjected to long-term oestrogen deprivation, the addition of low-dose oestrogen therapy paradoxically induces apoptosis, and a complete reversal of resistance to endocrine therapy has been shown (Masamura et al, 1995; Osipo et al, 2005).

In the clinical setting, Howell et al (1992), found in women with MBC that after prior lack of response to tamoxifen (19%) or after adjuvant tamoxifen followed by disease progression, 14% of patients experience stable disease (or disease response) when tamoxifen was stopped and the patients were observed. Although, it appears paradoxical that cessation of ‘anti-oestrogen’ therapy and subsequent exposure of tumour cells once again to normal circulating levels of oestrogens should induce disease regression, preclinical models have provided supportive evidence that this happens on occasion. Disease responses have also been seen in this setting of acquired disease resistance when therapeutic doses of oestrogens have been given to patients with such disease progression (Ingle, 2002; Ellis et al, 2009).

Therefore, in the clinical setting, the withdrawal of the AI at the time of tumour progression while on treatment could result in re-sensitivity to oestrogen, albeit to just normal menopausal levels. Thus moving immediately from one endocrine therapy to another in hormone-receptor-positive MBC, may be the wrong strategy in some patients.

We are aware that this is a small phase 2 study, however, our results are strongly suggestive of a withdrawal response in some patients: half of the patients with MBC progressing on AI therapy experienced disease stabilisation after withdrawal of the AI therapy. Two patients had remained progression free after 60 months (in one patient PFS was 73 months and had three sites of metastatic disease (bone, lymph nodes and contralateral breast). The other patient PFS was 66 months and had only one site of bone metastasis. Both patients had and ECOG of 0 and none of them received prior metastatic chemotherapy regimens). The PFS observed in our trial was of 4 months, which compares favourably to the PFS achieved with alternative options such as exemestane or fulvestrant, with a median PFS duration from 4 to 7 months, respectively, for patients with hormone-receptor-positive breast cancer (Chia et al, 2008; Di Leo, et al, 2011). We acknowledge a number of limitations of our study: it is a single-arm trial with a high likelihood of selection bias in favour of patients with more favourable tumour biology. With this caveat in mind unlike in the tamoxifen withdrawal trial published by Howell et al, we saw no objective disease remissions. Other possible selection biases include the following: patients were excluded from trial entry if their treating physician felt that they needed palliative chemotherapy, which potentially limited the study population to patients with slow-growing disease; the median age of our patients was 63 and many had bone-only disease (29%).

To our knowledge, this is the first and only prospective study of AI therapy withdrawal, which needs to be further, explored in a randomised trial. If the trend of our trial is confirmed this strategy could lead to a well tolerated, inexpensive and effective therapeutic option for selected patients with slowly progressing disease on endocrine therapy.