CDK 4/6 Inhibitor Palbociclib (PD0332991) in Rb⁺ Advanced Breast Cancer: Phase II Activity, Safety, and Predictive Biomarker Assessment

The G₁–S checkpoint is frequently dysregulated in breast cancer, leading to unchecked cellular proliferation. These include cyclin D1 overexpression or amplification, and loss of inhibitors p27 or p16 (1–3). Such tumors demonstrate increased proliferative indices, as measured by elevated levels of Ki-67 and increased phosphorylation of retinoblastoma (Rb) protein (4–7).

Palbociclib (PD0332991) is an oral inhibitor of the cyclin-dependent kinases (CDKs) 4 and 6 (8), preventing cell-cycle progression from G₁–S phase, thereby inhibiting cell proliferation and cellular DNA synthesis (9). Activity, resulting in blockade of Rb phosphorylation and G₁ arrest, was seen in Rb-positive tumor cell lines but not in Rb-negative MDA-MB-468 xenografts. In MDA-MB-435 xenografts, tumor growth suppression was shown to downregulate E2F-regulated genes (CDC2, CCNE2, TK1, TOP2A). MMTV-c-neu mice treated with palbociclib showed marked reduction in tumor volume and improved median survival compared with untreated controls (10), which was highly associated with presence of intact Rb and loss of p16.

In phase I, palbociclib was well tolerated and active, leading to a recommended phase II dose of 125 mg daily on a 3-week-on/1-week-off (3/1) schedule (11). Neutropenia was the dose-limiting toxicity; grade 3 neutropenia and anemia were seen in 12% and 7% of patients, respectively. The most common other toxicities were fatigue (34%), nausea (24%), constipation (17%), and vomiting (20%). While there were no partial responses (PR) in phase I, stable disease (SD) was seen in 35% of patients after 2 cycles, 27% of patients after 4 cycles and 16% of patients after 10 cycles. Notably, a heavily pretreated patient with ER⁺/Her2⁻ breast cancer achieved stable disease for 17 cycles.

On the basis of these findings, we conducted a phase II trial of palbociclib in patients with Rb-positive, advanced breast cancer. Primary endpoints were disease response and tolerability; secondary endpoints included progression-free survival (PFS) and biomarker assessment to determine whether Rb localization, Ki-67 index, p16 loss, or CCND1 amplification were associated with response.

This was a single-institution, open-label, nonrandomized, single-arm trial (UPCC03909; NCT01037790) that included breast, testis, colon, and gastric cancers. The current report focuses exclusively on the breast cancer cohort. The trial was approved by the Institutional Review Board at the University of Pennsylvania and the Scientific Review Committee of the Abramson Cancer Center. P. O'Dwyer served as both the Principal Investigator and study sponsor for this investigator-initiated trial. Pfizer provided funding and palbociclib, but did not participate in data collection, analysis, or writing of the article.

Eligible patients had histologically confirmed metastatic breast cancer staining positive for Rb (>1+ staining intensity as defined below) in primary or metastatic tumor, at least one measureable lesion by RECIST (V1.0; ref. 12) and adequate organ and bone marrow function. There was no limit to the number of prior therapies allowed. Required washout was 21 days for prior cytotoxic chemotherapy and 28 days for prior investigational agents. Stable, previously treated brain metastases were allowed. Uncontrolled intercurrent illness, a baseline QTcB interval >470 ms, pregnancy, breastfeeding, or human immunodeficiency virus infection were exclusionary.

Palbociclib was administered at 125 mg orally once daily, on days 1 to 21 of a 28-day cycle, continuing until disease progression or unacceptable toxicity. Toxicity was assessed using National Cancer Institute Common Terminology Criteria for Adverse Events Version 3.0 (CTCAE V 3.0) in cycle 1 on days 1, 8, 15, and 21 and then on day 1 of subsequent cycles. In the event of grade 3/4 toxicities, palbociclib was withheld until resolution to <grade 2, and resumed at the next lower dose. Dose levels for dose reduction were 100, 75, and 50 mg; patients requiring reduction below 50 mg were discontinued. Supportive care was allowed at the investigator's discretion, but strong inducers or inhibitors of CYP3A4 were prohibited.

Tumor assessments for response were conducted after every 2 cycles, using RECIST Version 1.0, by the RECIST core faculty of the Abramson Cancer Center blinded to patient identification and dose. Assessments were reduced to every 3 cycles for patients on treatment >18 months.

Archival tumor tissue from either the breast primary tumor or a metastatic lesion was required, and stained for Rb, Ki-67, and p16 by immunohistochemistry. Rb was stained with a 1:50 dilution of antibody MS-107-P, clone 1F8 (Neomarkers/LabVision; ref. 13). Ki-67 was stained with a 1:100 dilution of antibody MIB-1 (DAKO), against the MK167 (FHA domain) interacting nucleolar phosphoprotein (14). p16 was stained with a 1:200 dilution of antibody MS218 (Neomarkers; ref. 15). CCND1 amplification was assessed by FISH utilizing LSI Cyclin D1 SpectrumOrange/CEP11 SpectrumGreen (Vysis; ref. 16).

For each of the immunohistochemically stained markers, 2 independent pathologists determined subcellular location, staining intensity, and percentage of positively staining cells; discordance was resolved through adjudication. For Rb, categorical variables were created a priori by the method used by Shi and colleagues (13) based on a combination of percent tumor staining and intensity scores as follows: negative = 0 or 1+ intensity; equivocal = 2+ intensity or 3+ intensity in <30%; positive = 3+ staining in ≥30%, and separately assessed for nuclear and cytoplasmic compartments. p16 was categorized as 0%, 1–24%, 25–75% and >75% per previous reports (17, 18). Ki-67 staining was dichotomized at the 10% staining level (19). Amplification of CCND1 by FISH was defined by an 11q13:CEP ratio ≥2.2 (16).

We defined tumor responses as “complete response” (CR), “partial response” (PR), stable disease (SD), and progressive disease (PD) per RECIST 1.0 criteria (12). Because of the cytostatic mechanism of action of palbociclib, SD was added to CR and PR in the definition of “clinical benefit” (CB) if SD lasted ≥6 months. A Simon two-stage design (20) was employed, with a clinical benefit rate of ≥15% (at least 1 response in 15 patients) in stage 1 required to move on to stage 2. On the basis of our observations in stage 1 of activity exclusively in HR⁺ patients, we set our primary goal in stage 2 to estimate the response rate specifically in HR⁺ disease, stratifying on this marker. A sample size of 30 response-evaluable patients with HR⁺ disease was required to confirm a clinical benefit rate between 5% (p0) and 25% (p1; α and β both set at 0.10). We used Fisher exact test to assess association of categorical variables. PFS was defined as the time from the date of the initial palbociclib treatment to the date of death or date of progression (whichever occurred first), and PFS curves were planned a priori for the study population overall, by receptor subset (HR⁺, Her2⁺, and ER⁻/Her2⁻), and by prior treatments. Patients who discontinued due to toxicity were censored at the date of study discontinuation. Log-rank tests were conducted to assess effects of receptor subsets and prior treatments on PFS. Cox proportional hazards models were used to assess the effects of biomarkers on PFS. We estimated the effects of predictors on the probability of response by logistic regression. All analyses were conducted in STATA 13 (StataCorp).

Between April 2010 and March 2013, 128 patients with histologically confirmed breast cancer consented to screening for tumor Rb expression. Of these, 115 were Rb-positive, 5 were Rb-negative, and 8 had no tissue sample available. Thirty-seven patients who met other eligibility criteria subsequently provided informed consent and received treatment. Patient characteristics (Table 1) show a median age of 59 years (range 39–88). Eighty-four percent of tumors were HR⁺/Her2⁻, 5% HR⁺/Her2⁺, and 11% HR⁻/Her2⁻ (triple negative). This was also a heavily pretreated group, with a median of two prior chemotherapy regimens for metastatic disease, and up to 13 different regimens. Among those with HR⁺ tumors, the majority had also received prior hormone therapy, with 70% having at least two prior regimens.

Among the first 15 patients with breast cancer enrolled on the protocol, 2 had PR, meeting our prespecified criteria for continued enrollment. We completed full enrollment within the HR⁺ subset, but halted enrollment to the triple-negative cohort early because of observed rapid progression of enrolled patients and other treatment options available for this population.

Overall, CB was achieved in 7 of 37 patients, all in the 33 HR⁺ patients, for a clinical benefit rate (CBR) of 19% in the overall trial cohort and 21% in those who were HR⁺ (Table 2). Among HR⁺ patients, CB was seen exclusively in those patients who had already progressed through ≥2 prior lines of hormonal therapy (7/24, or 29%). With regard to prior chemotherapy, CBR was greater among patients with fewer than 2 prior regimens for metastatic disease compared with those who were more heavily treated (44% vs. 11% P = 0.045). Two patients in the HR⁺ subset also had tumors with Her2 overexpression. One patient experienced a PR; one patient had SD lasting 5 months. Neither Her2⁺ patient received concurrent Her2-directed therapy.

Figure 1 shows the waterfall plot for each patient's best tumor response by RECIST. Tumor shrinkage was greatest in a patient with HR⁺/Her2⁺ disease who was receiving no concurrent hormonal or Her2-directed therapy. To further illustrate the variability in duration of stable disease, the number of cycles received by both response assignment and receptor group is shown in Supplementary Fig. S1.

Figure 2 shows the Kaplan–Meier curves for PFS. The median PFS for the group overall (Fig. 2A) was 3.7 months (95% CI 1.9–5.1). Median PFS was 3.8 months (1.9–5.8) for patients with HR⁺/Her2⁻ disease, 5.1 months (5.1–∞) for HR⁺/Her2⁺ disease and 1.5 months (0.62–∞) for HR⁻/Her2⁻ disease. Patients with HR⁺ breast cancer (Fig. 2B) had significantly longer PFS compared with that of the HR⁻ group (4.5 months vs. 1.5 months, P = 0.03). Stratifying on degree of prior treatment (Fig. 2C and D), those with HR⁺ tumors and ≥2 prior lines of hormone therapy had significantly longer PFS (5 months; 95% CI, 2–6); compared with those who received fewer than 2 prior lines of hormone therapy (2 months; 95% CI, 1–4; P = 0.023); there was no difference in median PFS in patients with <2 prior cytotoxic regimens compared with those with ≥2 (median 5 months; 95% CI, 2–6; P = 0.114).

The incidence of adverse events by grade is shown in Table 3. One patient chose to discontinue study therapy after 2 cycles because of grade 2 fatigue. This patient had undergone three prior chemotherapy regimens for metastatic disease, but did not experience grade 3 or 4 neutropenia, anemia, or thrombocytopenia while on study therapy. No other patient was discontinued for toxicity and there were no treatment-related deaths.

Overall, there were 56 grade 3 events and 3 grade 4 adverse events, all due to myelosuppression. Nineteen patients (51%) experienced grade 3/4 neutropenia; 8 (22%) experienced grade 3/4 thrombocytopenia. Of those with grade 3/4 anemia (2 patients), 1 had grade 1 anemia at baseline. Grade 3/4 neutropenia was relatively isolated; only 6 of 19 (32%) had concurrent grade 3/4 anemia or thrombocytopenia. It was also largely uncomplicated, with only one episode of fever and sepsis in a neutropenic patient who was experiencing active disease progression simultaneously.

Myelosuppression, particularly neutropenia, led to dose interruptions and reductions (Table 4). Forty-six percent of all dose modifications were due to neutropenia, mostly grade 3. Among those with dose reductions, a significant fraction (19%) had more than one dose reduction for myelosuppression, with 7 patients being reduced to 50 or 75 mg daily. Of the 19 dose reductions, 14 (74%) occurred at the start of cycle 2, while only 3 occurred during cycle 1 (at the 2-week CBC), and just 1 occurred beyond cycle 2 (at cycle 5). These dose reductions did not appear to adversely affect response: among the 7 patients who experienced clinical benefit, only 2 remained at the 125 mg dose; 2 were reduced to 100 mg, 2 were reduced to 75 mg, and 1 was reduced to 50 mg, all with continued response.

Representative sections for each of the biomarkers of interest, expression frequencies, and associations with odds of response and hazards of progression are shown in Supplementary Fig. S2 and Supplementary Table S1, respectively. None of the markers we evaluated were significantly associated with either clinical benefit or PFS, either at the a priori hypothesized cutoff points or in exploratory analyses of continuous data both in the overall and HR⁺ groups (21).

Palbociclib (PD0332991), an oral CDK4/6 inhibitor, demonstrates single-agent activity in this heavily pretreated population of patients with advanced breast cancer. The CBR was 21% for patients with HR⁺ disease, but rose to 29% among those who had progressed through at least 2 prior lines of hormonal therapy, suggesting substantial activity in the setting of acquired endocrine resistance. These patients also had prolonged PFS (5 months) compared with those patients who had had only one line of prior endocrine therapy. Response duration in second line hormone therapy is typically 3 to 4 months, as evidenced by the 3.7-month PFS seen with both exemestane and fulvestrant in the phase III EFECT trial (22) and the 2.8 month PFS for the exemestane control arm in BOLERO-2 (23). Thus, the level of activity we saw in patients demonstrating prior endocrine resistance is encouraging, and will be further tested to some extent in the PALOMA-3 trial, combining palbociclib with fulvestrant in patients with endocrine-resistant breast cancer.

Palbociclib does appear to improve PFS in the first-line setting when combined with endocrine therapy compared with endocrine therapy alone, as seen in the recently completed randomized phase II PALOMA-1 trial (24). Because our study included only 1 first-line patient, it is impossible to draw conclusions about the single agent activity of palbociclib in first-line patients who are typically endocrine sensitive.

The finding of responses in HR⁺/Her2⁺ disease, as seen in the prolonged responder in this study, is supported by preclinical data suggesting sensitivity of Her2⁺ cells to cell-cycle inhibition (25). The fact that this occurred in the absence of Her2-directed therapy suggests an alternative mechanism downstream of Her2 signaling. We enrolled only 4 patients with triple -negative disease, all of whom rapidly progressed on treatment. This, does not preclude the possibility that triple-negative disease could respond to palbociclib; combinations with chemotherapy and other targeted therapies are currently under investigation (26). Like the PALOMA-1 trial, our study did not identify any associations with hypothesized cell-cycle alterations, though we were unable to assess target suppression in the absence of on-treatment tumor biopsies.

Palbociclib was extremely well tolerated in this trial, and the absence of common systemic treatment-related symptoms such as alopecia, nausea, diarrhea, rash, or pain in our patients was notable, particularly in the setting of response. The emergence of uncomplicated grade 3/4 neutropenia suggests that the mechanism of myelosuppression with palbociclib may differ from that of traditional cytotoxic chemotherapy. The rarity of neutropenic fever/infection suggests that bone marrow progenitors, suppressed during treatment, may still be functional when faced with an infectious challenge, as preclinical studies have suggested (27). Sharpless and colleagues demonstrated that exposure of hematopoietic stem cells in vitro to PD0332991 reversibly decreased proliferation, but did not decrease total marrow cellularity, or alter apoptosis or viability of hematopoietic progenitor cells. These findings are corroborated by our clinical observation that short interruption followed by dose reduction was effective in restoring normal neutrophil counts. Thus, though the incidence of neutropenia is high, the relative paucity of fever and infection precludes this from being a major safety concern. It is not clear to what extent dose reductions altered response to therapy; however, response was seen at doses as low at 50 mg daily.

Recently, results were reported for two other cdk4/6 inhibitors currently in development. A phase I trial of LEE011 (Novartis) in advanced Rb-positive solid tumors and lymphoma (CLEE011 × 2101/NCT01237236) reported at ASCO, 2014 (28). At an MTD of 900 mg/day on a 3/1 schedule, 29% of patients had grade 3/4 neutropenia, 10% had grade 3/4 thrombocytopenia, and 3% had grade 3/4 anemia, fatigue, and/or diarrhea. Abemiclib (LY2835219, Lilly) was tested in a multitumor phase I/II trial, with breast cancer cohort results reported at AACR, 2014 (29). At the MTD of 200 mg twice daily on a continuous 28-day cycle, 19% of patients had grade 3 neutropenia and 10.6% had grade 3 thrombocytopenia. However, the most common toxicity was diarrhea, which occurred in 43%, 17%, and 6.4% of patients at grades 1, 2, and 3, respectively. Nausea, fatigue, and vomiting additionally occurred in at least 15% of patients. There was one episode of febrile neutropenia. Among this heavily pretreated cohort (median 7 lines of prior therapy), 19% of patients experienced PR and 29.8% experienced ≥6 months SD. PR and 6-month SD rates increased to 25% and 36%, respectively, among women with HR⁺ disease. The median PFS was 9.1 months with 18 patients still on study. The reasons for the increased single-agent response rate with this agent relative to LEE011 and palbociclib are not clear, but could be a result of a differing mechanism of action or a modulatory pharmacodynamic effect of continuous dosing. Continuous dosing has not been tested with palbociclib to date, precluding direct comparison of these agents.

Given the sample size constraints of this trial, we have interpreted our data with caution. We have reported only the prespecified analyses for HR⁺ patients and comparison by prior therapy. We limited biomarker analyses to biologically hypothesized common cell-cycle changes at prespecified cutoff points and exploratory analysis of continuous variables, but because of the small samples, these findings in particular must be viewed as hypothesis-generating only, and warrant further study. Furthermore, other biomarkers may be important to predicting response to therapy. Recent data from the Engelman laboratory (30), for example, have suggested that PI3K alterations could sensitize tumors to the effects of cdk 4/6 inhibition and other alterations, such as Rb mutation or L0H could be important as well (31).

In summary, palbociclib is an active single agent in metastatic breast cancer, primarily in endocrine-resistant, HR⁺ disease. These findings, as well as the promising findings in PALOMA-1 and its favorable side-effect profile have led to further development of palbociclib in breast cancer. Ongoing trials of this agent in HR⁺ disease with endocrine therapy (PALOMA-2 and -3) and in combination with chemotherapy will further elucidate the role of palbociclib in the treatment of breast cancer.

A. DeMichele is a consultant/advisory board member for Pfizer. K. Gogineni reports receiving a Medical Academic Partnership Research in Bioethics Grant from Pfizer. K. Fox is a consultant/advisory board member for Novartis. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A. DeMichele, P. O'Dwyer

Development of methodology: A. DeMichele, A.S. Clark, M. Feldman, P. Zhang, P. O'Dwyer

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. DeMichele, A.S. Clark, M. Gallagher, P. Lal, M. Feldman, P. Zhang, C. Colameco, D. Lewis, M. Langer, N. Goodman, S. Domchek, K. Gogineni, K. Fox, P. O'Dwyer

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. DeMichele, A.S. Clark, K.S. Tan, D.F. Heitjan, C. Colameco, P. O'Dwyer

Writing, review, and/or revision of the manuscript: A. DeMichele, A.S. Clark, K.S. Tan, D.F. Heitjan, P. Lal, M. Feldman, S. Domchek, M. Rosen, P. O'Dwyer

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. DeMichele, A.S. Clark, D.F. Heitjan, K. Gramlich, P. Zhang, D. Lewis, N. Goodman, M. Rosen

Study supervision: A. DeMichele, M. Feldman, N. Goodman, P. O'Dwyer

This study was supported by the Penn–Pfizer Alliance, Expedition Inspiration Fund for Breast Cancer Research.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.