TBCRC 035: randomized phase II pharmacodynamic study of standard and reduced-dose palbociclib with endocrine therapy in hormone receptor (HR)-positive previously treated metastatic breast cancer

Abstract

Background

The cyclin-dependent kinase 4/6 inhibitor palbociclib induces neutropenia, resulting in dose delays and reductions.

Patients and methods

This multicenter phase II trial randomly assigned patients with metastatic hormone receptor-positive, human epidermal growth factor receptor 2-negative breast cancer to receive palbociclib 100 mg versus 125 mg with fulvestrant or tamoxifen; cross-over was allowed. Baseline and on-treatment skin and tumor biopsies were analyzed for expression of phosphorylated retinoblastoma protein (pRb), total Rb, and Ki-67. Baseline circulating tumor DNA was collected. The primary endpoint was grade ≥3 neutropenia.

Results

Seventy patients with a median of 3 prior treatment lines (range 0-6) were enrolled; 36 received 100 mg and 34 received 125 mg. Grade ≥3 neutropenia occurred in 12 (33%) and 19 (56%) patients in the 100 versus 125 mg groups, respectively (P = 0.04). Median progression-free survival (PFS) was 6.28 versus 9.28 months for 100 versus 125 mg, respectively (hazard ratio 1.697, 95% confidence interval 0.973-2.96, P = 0.0585). Five patients crossed over to 125 mg; two patients were treated for >12 months. Tumor and skin pRb and Ki-67 decreased on treatment, with similar percent changes across palbociclib dose (tumor pRb: −2.16 and −4.64, P = 0.897; tumor Ki-67: −10.27 and −6.7, P = 0.437 for 100 mg and 125 mg, respectively; skin pRb and Ki-67 P < 0.02 at both doses). Baseline mutations in PIK3CA and TP53, and higher baseline Ki-67 were associated with shorter PFS. Changes in pRb and Rb were not associated with PFS.

Conclusions

Palbociclib 100 mg was associated with a reduced incidence of ≥grade 3 neutropenia. While PFS was numerically lower at 100 mg, the difference was not statistically significant and was limited by sample size and complicated by varying degrees of pre-treatment. Analysis of skin and tumor pRb or Ki-67 demonstrated robust molecular response at both doses. PIK3CA and TP53 mutations and higher baseline Ki-67 were associated with inferior clinical outcome.

Highlights

• •A lower starting dose of palbociclib 100 mg caused less grade ≥3 neutropenia versus 125 mg (33% versus 56%, P = 0.04).
• •Median PFS was longer but not significant at 125 mg (9.3 versus 6.3 months) in this small, pre-treated cohort.
• •Tumor and skin pRb and Ki-67 decreased similarly at both doses indicating on-target effect.
• •Baseline PIK3CA and TP53 mutations, and high tumor Ki-67 predicted shorter PFS.
• •Dose optimization efforts incorporating biomarkers are imperative to minimize toxicity without compromising efficacy.

Introduction

The addition of a cyclin-dependent kinase 4/6 inhibitor (CDK4/6i) to endocrine therapy (ET) is the standard of care for patients with metastatic hormone receptor (HR)-positive/human epidermal growth factor receptor 2 (HER2)-negative breast cancer in the first line and pretreated settings, with multiple randomized phase III trials showing progression-free survival (PFS) and in some cases overall survival (OS) benefit compared with ET alone.1, 2, 3, 4, 5, 6, 7, 8 The CDK4/6is prevent phosphorylation of the tumor suppressor retinoblastoma protein (Rb), thereby inhibiting cell cycle progression from G1 to S phase and cell proliferation.^9,10

Standard of care CDK4/6is utilized in the advanced breast cancer setting include palbociclib, ribociclib, and abemaciclib. Palbociclib was the first CDK4/6i to be approved by the United States Food and Drug Administration (FDA) based on the PALOMA trials.^1,2,4 Although generally well tolerated, specific toxicities are observed, notably neutropenia with palbociclib and ribociclib, leading to dose holds, dose reductions, and even discontinuations. Phase III trials of palbociclib reported rates of grade ≥3 neutropenia between 50% and 60% with dose reductions in over 30% of patients. Febrile neutropenia is rare, occurring in ∼1% of patients.^1,4,11 While dose reductions are common in clinical practice, prospective trials of dose-reduced palbociclib are lacking; further understanding may help direct clinical practice.

We conducted a phase II randomized trial evaluating palbociclib at a starting dose of either 125 mg or 100 mg with provider choice ET (tamoxifen or fulvestrant). We hypothesized that palbociclib administered at 100 mg would result in lower rates of grade ≥3 neutropenia without significant impact on PFS. We further hypothesized that the 100 mg dose of palbociclib would have a similar impact on pharmacodynamic biomarkers such as phosphorylated retinoblastoma protein (pRb) and Ki-67 expression in skin and tumor tissue compared with the 125 mg dose. Alterations in specific genes in circulating tumor DNA (ctDNA) were analyzed at baseline to assess the impact on outcome in both groups.

Patients and methods

Study design, eligibility, and procedures

This was a prospective randomized phase II, open-label, multicenter trial that enrolled at 11 cancer centers between 11 February 2016 and 27 November 2018 within the Translational Breast Cancer Research Consortium. Eligible patients had histologically proven metastatic breast cancer, estrogen receptor and/or progesterone receptor-positive (HR -positive) disease as defined by ≥1% positively stained cells, and HER2-negative disease (immunohistochemistry score 0/1+ or without gene amplification by in situ hybridization). Pre-menopausal women were required to receive ovarian suppression. All patients had measurable disease by RECIST 1.1 criteria and the ability to have a skin and tumor biopsy.

Eligible patients had three or lesser prior chemotherapy lines and any number of ET lines for advanced disease. Adjuvant chemotherapy within 1 year of recurrence counted as one line. Prior CDK4/6i was not allowed; prior aromatase inhibitor, fulvestrant, and tamoxifen were permitted. Adequate organ function was required.

Patients were randomly assigned in a 1 : 1 ratio to receive palbociclib 125 mg or 100 mg orally on days 1-21 of a 28-day cycle, with investigator’s choice of tamoxifen 20 mg daily or fulvestrant 500 mg intramuscular (days 1 and 15 of cycle 1, then day 1 of subsequent cycles). Randomization was stratified by the choice of ET. Absolute neutrophil count >1000/μl was required on day 28 to start the next cycle; palbociclib could be delayed up to 3 weeks for neutropenia, after which discontinuation was mandated. Dose reductions were advised for prolonged grade >2 or any grade ≥3 toxicities but not below 75 mg. Patients progressing on 100 mg could cross over to 125 mg until further progression. Response was assessed by RECIST 1.1 imaging at baseline and every 8 weeks, with clinical assessments every 4 weeks. Treatment continued until progression, toxicity, withdrawal, or investigator’s decision. Baseline and cycle 1 (days 14-21) biopsies were obtained, and ctDNA was collected at baseline.

The study protocol was reviewed and approved by the FDA and each institutional review board (IND 124869; NCT02384239). All patients provided written informed consent. All study procedures were carried out in accordance with the Declaration of Helsinki.

Outcome analysis

The primary objective of this study was the incidence of grade ≥3 neutropenia as defined by Common Terminology Criteria for Adverse Events version 4 in patients taking palbociclib at either 100 mg or 125 mg with ET. Secondary outcomes included PFS by dose, objective response rate (ORR), clinical benefit rate (CBR) defined as a complete response, partial response or stable disease for at least 6 months, inhibition of Rb phosphorylation and Ki-67 expression in tumor and skin samples, toxicity of palbociclib, and PFS in those with ctDNA mutations.

Tissue and blood analysis

Baseline and on-treatment skin punch and tumor core biopsies were formalin-fixed, paraffin-embedded, sectioned (5 μM), and stained for pRb (S780) (Cell Signaling Technology #9308, Danvers, MA), total Rb (clone 3C8, LS-3414, Abnova, Taipei City, Taiwan), and Ki-67 (clone MIB-1, #M7240, Agilent Dako, Santa Clara, CA) on the Leica BOND IHC platform using 3, 3-diaminobenzidine or polymer red detection (to reduce melanin background). Slides were scanned on the Aperio platform¹² (Deer Park, IL)and percent marker-positive cells quantified. Tissue analyses were carried out at the Dana-Farber Cancer Institute. Whole blood collected at baseline was processed for plasma and ctDNA analyzed with the SafeSEQ Breast Cancer panel (Sysmex Inostics, Baltimore, MD; cut-off 0.05% mutant: wild-type alleles),¹³ targeting AKT1, ERBB2, KRAS, PIK3CA, and TP53. Correlative ctDNA analysis by OncoBEAM (Sysmex Inostics) droplet PCR showed excellent concordance for PIK3CA, ESR1, and AKT1 mutations (R² = 0.9819).¹⁴ All ctDNA assays were carried out by Sysmex Inostics.

Statistical analysis

We tested the incidence of grade ≥3 neutropenia against the null that rates would be <63.6%, based on PALOMA-1.² A single-stage binomial exact test (one-sided α = 0.05) was used per arm. With 35 patients per dose level, the power was 78% to reject the null if the true rate was 43.6% (20% reduction). Thus, 70 patients (35 per dose level) were planned. Analyses included all patients receiving one or more dose of palbociclib. Adverse events were reported at each patient’s highest grade. For the primary endpoint, the null was rejected if ≤17/35 patients (<49%) had grade ≥3 neutropenia. Observed rates were reported with two-sided 90% exact confidence intervals (CIs).

The secondary endpoint of PFS was calculated using the Kaplan–Meier method with 95% confidence bands computed using Greenwood’s formula. Objective response and clinical benefit were calculated with a 95% exact binomial CI. Hazard ratios were calculated using the Cox proportional hazards model.

Descriptive statistics summarized total Rb, pRb, and Ki-67 at baseline and D14-21. Absolute changes from baseline were tested with paired Student’s t-test. Differential pRb inhibition between palbociclib 100 mg and 125 mg was evaluated using a general linear model with baseline as a covariate (two-sample t-test) and reported with 95% CIs. Correlations between baseline skin/tumor measures and changes from baseline were assessed by Pearson coefficients. Associations between pRb inhibition in skin/tumor and PFS were analyzed using the Cox model, with continuous factors assessed by C-index and varying endpoint partitions.¹⁵ Correlations of total Rb, pRb, and Ki-67 at baseline with PFS were evaluated by Spearman’s rank test. Plasma ctDNA was assessed at baseline using descriptive statistics. PFS for patients with and without ESR1, PIK3CA, and TP53 mutations were assessed using the Kaplan–Meier method with 90% confidence bands computed using Greenwood’s formula. Hazard ratios for PFS were calculated using the Cox proportional hazards model.

Results

Between 11 February 2016 and 27 November 2018, 70 patients were randomly allocated to the study at 11 centers (36 to the 100 mg group, 34 to the 125 mg group). Patient demographics and clinical characteristics are outlined in Table 1. Most received fulvestrant (100 mg: 22, 61%; 125 mg: 21, 62%). Five patients in the 100 mg group had de novo metastatic disease, all receiving tamoxifen, compared with none in the 125 mg group. Most patients had received a prior aromatase inhibitor (100 mg: 26, 72%; 125 mg: 27, 79%). Median number of prior lines of therapy in the metastatic setting was 3 (range 0-6) with one median line of chemotherapy and one median line of ET.

Table 1.

Patient demographics and clinical characteristics

	Palbociclib 100 mg	Palbociclib 125 mg	All
No. of patients	36	34	70
Median age, years (range)	60 (39-88)	58 (42-82)	59 (39-88)
Gender, n (%)
Female	36 (100%)	33 (97%)	69 (99%)
Male	0	1 (3%)	1 (1%)
Race, n (%)
White	29 (81%)	26 (76%)	55 (79%)
Black or African American	2 (6%)	1 (3%)	3 (4%)
Native Hawaiian or other Pacific Islander	0 (0%)	1 (3%)	1 (1%)
Asian	1 (3%)	2 (6%)	3 (4%)
American Indian or Alaska Native	0	0	0
Unknown	4 (11%)	4 (12%)	8 (12%)
Ethnicity, n (%)
Hispanic or Latino	1 (3%)	0	1 (1%)
Non-Hispanic	33 (92%)	33 (97%)	66 (94)
Unknown	2 (6%)	1 (3%)	3 (4%)
ECOG performance status, n (%)
0	4 (11%)	9 (26%)	13 (19%)
1	32 (89%)	23 (68%)	55 (79%)
Unknown	0 (0%)	2 (6%)	2 (3%)
De novo stage IV disease, n (%)	5 (14%)	0	5 (7%)
Disease sites, n (%)
Visceral disease	20 (56%)	19 (56%)	39 (56%)
Bone only	6 (17%)	9 (26%)	15 (21%)
Estrogen receptor expression in screening tumor biopsy, n (%)
<10%	6 (17%)	5 (15%)	11 (16%)
>10%	30 (83%)	29 (85%)	59 (85%)
Progesterone receptor expression in screening tumor biopsy, n (%)
<10%	16 (44%)	21 (62%)	37 (53%)
>10%	20 (56%)	13 (38%)	33 (47%)
Endocrine partner, n (%)
Tamoxifen	14 (39%)	13 (38%)	27 (39%)
Prior tamoxifen in those receiving tamoxifen	3 (8%)	6 (18%)	9 (13%)
Fulvestrant	22 (61%)	21 (62%)	43 (61%)
Prior fulvestrant in those receiving fulvestrant	3 (8%)	6 (18%)	9 (13%)
Prior aromatase inhibitor in any setting, n (%)	26 (72%)	27 (79%)	53 (76%)
Median lines of therapy in the metastatic setting (range)	3 (0-6)	2.5 (0-6)	3 (0-6)
Median lines of chemotherapy	1 (0-4)	1 (0-3)	1 (0-4)
Median lines of hormone therapy	1 (0-4)	1 (0-3)	1 (0-3)
Adjuvant therapy, n (%)
Chemotherapy	17 (48%)	22 (65%)	39 (56%)
Endocrine therapy	22 (61%)	27 (79%)	49 (70%)

ECOG, Eastern Cooperative Oncology Group.

Safety

Seventy patients were assessable for the primary endpoint with 12 (33%) in the 100 mg group and 19 (56%) in the 125 mg group experiencing grade ≥3 neutropenia (Table 2), with significantly less grade ≥3 neutropenia in patients receiving 100 mg versus 125 mg (P = 0.04). Eight patients in the 100 mg group required dose delays due to neutropenia compared with 11 in the 125 mg group. Other adverse events of any grade seen in >10% of the total study population included fatigue (57%, n = 40 total, n = 19 at 100 mg, n = 21 at 125 mg), nausea (21%, n = 15 total, n = 6 at 100 mg, n = 9 at 125 mg), and mucositis (17%, n = 12 total, n = 5 at 100 mg, n = 7 at 125 mg). Two patients had serious adverse events attributed to palbociclib at the 100 mg dose, including grade 2 limb edema and grade 3 gastrointestinal infection.

Table 2.

Adverse events

	Palbociclib 100 mg n= 36		Palbociclib 125 mg n = 34		All N = 70
	Any grade, n (%)	G ≥3, n (%)	Any grade, n (%)	G ≥3, n (%)	Any grade, n (%)	G ≥3, n (%)
Adverse event with >10% incidence in total population
Neutropenia	15 (42%)	12 (33%)	23 (68%)	19 (56%)	38 (54%)	31 (44%)
Fatigue	19 (53%)	0 (0%)	21 (62%)	1 (3%)	40 (57%)	1 (1%)
Nausea	6 (17%)	0 (0%)	9 (26%)	1 (3%)	15 (21%)	1 (1%)
Mucositis	5 (14%)	0 (0%)	7 (21%)	0 (0%)	12 (17%)	0 (0%)
Anemia	2 (6%)	0 (0%)	6 (18%)	1 (3%)	8 (11%)	1 (1%)
Anorexia	4 (11%)	0 (0%)	4 (12%)	0 (0%)	8 (11%)	0 (0%)
Thrombocytopenia	3 (8%)	1 (3%)	5 (15%)	0 (0%)	8 (11%)	1 (1%)
Upper respiratory infection	4 (11%)	0 (0%)	4 (12%)	0 (0%)	8 (11%)	0 (0%)
Alopecia	2 (6%)	0 (0%)	5 (15%)	0 (0%)	7 (10%)	0 (0%)
Vomiting	2 (6%)	0 (0%)	5 (15%)	1 (3%)	7 (10%)	1 (1%)

G, grade.

Patients received a median of 5.5 months of palbociclib in the 100 mg group and 5.7 months in the 125 mg group (Supplementary Table S1, available at https://doi.org/10.1016/j.esmoop.2026.106063). In the 100 mg group, six patients (17%) were dose reduced to 75 mg (Supplementary Table S1, available at https://doi.org/10.1016/j.esmoop.2026.106063). An additional five patients who experienced grade ≥3 neutropenia who did not undergo dose reduction had a total of one dose delay. In the 125 mg group, 16 patients (47%) were dose reduced; 13 patients (38%) were reduced to 100 mg and 3 patients (9%) were reduced a second time to 75 mg. Four patients in each group discontinued treatment due to adverse events related to palbociclib (two due to neutropenia, one due to fatigue, two due to limb edema, two due to arthralgias, one due to elevated liver enzymes). The median time from treatment start to first dose reduction in the 100 mg group was 4.1 weeks and in the 125 mg group was 4.3 weeks. The median number of dose delays in both groups was 1 (range 1-4), with 10 patients experiencing dose delays (three patients in 100 mg group, seven in 125 mg group). Most dose reductions and delays in both groups were due to neutropenia (55 of 83 events, 66%); other reasons for dose reduction included fatigue, thrombocytopenia, gastrointestinal toxicity, infection, and rash.

Efficacy

Median PFS was 6.28 months [interquartile range (IQR) 1.88-13.26 months] in the 100 mg group and 9.28 months (IQR 1.97-19.38 months) in the 125 mg group (hazard ratio 1.697, 95% CI 0.973-2.960, P = 0.059; Figure 1). Three patients in each group experienced a partial response and 21 patients in each group experienced stable disease as the best response (Supplementary Table S2, available at https://doi.org/10.1016/j.esmoop.2026.106063). The CBR was similar between the two arms, at 67% in the 100 mg group compared with 75% in the 125 mg group (95% CI −0.2983 to 0.1317, P = 0.514). Twelve patients (33%) in the 100 mg group had progressive disease as best response compared with eight patients (25%) in the 125 mg group.

Figure 1.

Progression-free survival by treatment arm. Kaplan–Meier plot of progression-free survival (PFS) of patients randomly assigned to receive the 100 mg and 125 mg dose of palbociclib regardless of the endocrine therapy partner. Censoring (plus sign) and number of individuals at risk are shown below. Median PFS of patients were 6.28 and 9.28 months in the 100 and 125 mg arms, respectively (HR 1.697, 95% CI 0.973-2.96, P = 0.0585), and depicted in the table. CI, confidence interval; HR, hazard ratio.

Nine of 43 patients receiving fulvestrant on trial had received fulvestrant previously and 9 of 27 receiving tamoxifen on trial had received it previously. For 7 of these 18 patients, PFS was >6 months (median 13.25 months, range 11.01-18.66 months).

Five patients crossed over from 100 mg to 125 mg at the time of disease progression, with a median post-cross-over PFS of 1.75 months (0.25-19.25 months). Two of the five patients experienced prolonged PFS after cross-over (13.5 and 19.25 months, respectively) and total duration on study (20.25 and 47.25 months since study entry, respectively).

Correlative endpoints

Changes in pRb, total Rb, and Ki-67 were assessed between samples procured before treatment and during cycle 1 at both dose levels (Figure 2). Representative immunohistochemical stains are shown in Figure 2A. In the 27 available matched tumor specimens, there was a statistically significant decline in pRb (95% CI 27.7-70.0, P = 0.014) and Ki-67 (95% CI −2.8 to 0.73, P < 0.001; Figure 2B) from baseline to day 14-21 of cycle 1. Between the two dose levels, there was no significant difference in the change in percent positive cells for either pRb (−2.16 and −4.64 for the 100 and 125 mg doses, respectively) or Ki-67 (−10.27 and −6.7; Figure 2C). No significant changes in total Rb staining occurred between pre- and on-treatment samples. In the 66 available paired skin biopsies, a similar pattern was observed with a significant decline in pRb and Ki-67 in on-treatment compared with pre-treatment samples in both the 100 mg and 125 mg groups (all P < 0.001; CI for pRB 100 mg: −3.6 to −1.5; CI for pRB 125 mg: −4.4 to −2.5; CI for Ki-67 100 mg: −3.3 to −1.4; CI for Ki-67 125 mg: −4.1 to −2.4; Figure 2D). Consistent with the data in tumor specimens, there was no significant difference in the change in percent of cells staining positive for pRb (−2.85 versus −3.6 in the 100 and 125 mg groups) or Ki-67 (−2.9 versus −3.2 in the 100 and 125 mg groups,) between the two dose levels.

Figure 2.

Changes in Ki-67 and phosphorylated Rb in tumor tissue and skin. (A) Analysis of pharmacodynamic markers show marked reduction in on-treatment biopsy. Representative tumor biopsy specimens (patient 214-0143) obtained pre-treatment and on-treatment (C1D17) were stained using antibodies to phospho-retinoblastoma protein [pRb (S780)], total Rb, and Ki-67. (B) The box plot depicts the expression of Ki-67, total-Rb, and pRb in 27 paired tumor tissue samples, as determined by percentage of cells staining positive. Data are presented before and after treatment, regardless of palbociclib dose level or endocrine therapy partner. (C) The box plot depicts the change in the percent of cells staining positive for expression of Ki-67, total Rb, and pRb in tumor tissue samples between baseline and on-treatment for patients receiving both 100 mg and 125 mg of palbociclib. (D) The box plot depicts pRb and Ki-67 expression, as determined by the percentage of cells staining positive in 66 paired skin biopsies taken at baseline and between day 14 and 21 of the first cycle of palbociclib. Expression is shown pre-treatment and on-treatment at both dose levels of palbociclib. pRb, phosphorylated retinoblastoma protein; total-Rb, total retinoblastoma protein; Pre, pre-treatment; On-Tx, on-treatment

Among patients with paired biopsy samples, levels of total Rb, pRb, and Ki-67 were assessed to determine whether there were associations with PFS. These patients were also categorized based on the presence of ctDNA mutations at baseline (Supplementary Figure S1, available at https://doi.org/10.1016/j.esmoop.2026.106063). Baseline Ki-67 was negatively correlated with PFS with higher baseline tumor Ki-67 associated with shorter survival (Spearman’s correlation coefficient −0.4932, CI 0.0209-0.3236, P = 0.0273; Supplementary Figure S2A, available at https://doi.org/10.1016/j.esmoop.2026.106063). Neither the baseline levels of Rb or pRb nor the degree of reduction in pRb and Ki-67 on treatment compared with pre-treatment correlated with PFS.

Sixty-nine patients underwent ctDNA analysis at baseline with 52 patients having detectable alterations. The most commonly detected alterations were TP53 (30 patients, 43%), PIK3CA (27 patients, 39%), and ESR1 mutations (24 patients, 35%, Supplementary Table S3, available at https://doi.org/10.1016/j.esmoop.2026.106063). Analysis of PFS by ESR1 mutation showed no difference in those with and without mutations (6.5 versus 7.3 months, hazard ratio 0.880, 95% CI 0.502-1.540; Figure 3A). PIK3CA mutations were associated with inferior PFS (2.30 versus 9.14 months, hazard ratio 0.558, 95% CI 0.322-0.969; Figure 3B) as were TP53 mutations (PFS 3.72 versus 8.59 months, hazard ratio 0.562, 95% CI 0.323-0.976; Figure 3C). Further PFS analyses were carried out after categorizing patients with ctDNA into one of five groups: those with ESR1 mutations, those with PIK3CA/AKT mutations, those with TP53 mutations, those with HER2/KRAS mutations, and those with wild-type alleles in these genes (Supplementary Figure S3, available at https://doi.org/10.1016/j.esmoop.2026.106063). Patients with more than one of the defined mutations were categorized in separate groups, in accordance with their exact mutations, as outlined in Supplementary Figure S3, available at https://doi.org/10.1016/j.esmoop.2026.106063. Patients with simultaneous mutations in TP53, PIK3CA, and ESR1 (n = 8) demonstrated inferior PFS compared with those whose tumors were wild-type for all three genes (1.81 versus 7.17 months, hazard ratio 4.69, 95% CI 1.60-13.72; Figure 3D). In addition, patients with simultaneous mutations in both TP53 and PIK3CA compared with those who were wild-type for both genes demonstrated a trend toward inferior PFS (6.46 versus 7.17 months, hazard ratio 2.10, 95% CI 0.85-5.19; Figure 3E). Of note, four patients with PFS >6 months also harbored mutations in ctDNA: one patient with a mutation in TP53 (P152L), two with ESR1 (L536P, Y537S, and D538G), and one with ESR1 (Y537S and D538G) as well as PIK3CA (E542K) mutations. Higher Ki-67 was significantly correlated with higher pRb expression at baseline in those without mutations in TP53, ESR1, or PIK3CA, compared with those with mutations (Supplementary Figure S2B, available at https://doi.org/10.1016/j.esmoop.2026.106063).

Figure 3.

Progression-free survival by ctDNA mutation. (A) Progression-free survival (PFS) for patients with and without ESR1 mutations was calculated using the Kaplan–Meier method. Number at risk is depicted just below the x-axis for patients with and without ESR1 mutations. (B) PFS for patients with and without PIK3CA mutations was calculated using the Kaplan–Meier method. Number at risk is depicted just below the x-axis for patients with and without PIK3CA mutations. (C) PFS for patients with and without TP53 mutations was calculated using the Kaplan–Meier method. Number at risk is depicted just below the x-axis for patients with and without TP53 mutations. (D) PFS for patients with mutations in ESR1, PIK3CA, and TP53 mutations compared to those who are wild-type in all three genes was calculated using the Kaplan–Meier method. Number at risk is depicted just below the x-axis. (E) PFS for patients with mutations in PIK3CA and TP53 mutations compared to those who are wild-type in both genes was calculated using the Kaplan–Meier method. Number at risk is depicted just below the x-axis. CI, confidence interval; ctDNA, circulatory tumor DNA; HR, hazard ratio; mut, mutant; wt, wild-type.

Discussion

CDK4/6is in combination with ET have improved survival outcomes for patients with HR -positive/HER2-negative metastatic breast cancer in the first line and pretreated settings,1, 2, 3, 4^,7,8,16 yet neutropenia remains a challenge.1, 2, 3, 4, 5^,11 One single-arm trial of 54 patients evaluating a schedule of 5 days on, 2 days off drug showed an improved grade ≥3 neutropenia rate of 39% with median PFS of 19.75 months.¹⁷ Another retrospective review of real-world palbociclib dosing identified seven different changes to palbociclib dosing and scheduling and found that these strategies achieved comparable disease control to standard dosing.¹⁸ Finally, an exploratory analysis of dose reductions in the phase III PALOMA-3 trial, which combined fulvestrant with palbociclib at 125 mg, demonstrated no difference in PFS for patients undergoing dose reduction of palbociclib due to neutropenia compared with those who did not require a dose reduction.¹⁹

Our study prospectively evaluated the use of standard-dose palbociclib at 125 mg daily versus a lower initial dose of palbociclib at 100 mg daily in combination with ET and found that the 125 mg dose was associated with more frequent grade ≥3 neutropenia compared with the 100 mg dose (P = 0.037). The 17% rate of dose reduction in the 100 mg palbociclib cohort is much lower than the 30% rate of dose reduction seen in a pooled analysis of the PALOMA trials with the 125 mg starting dose.¹¹ Although patient reported outcomes (PROs) were not collected in our trial, the overall tolerability is consistent with favorable PROs reported in multiple larger phase III trials evaluating the addition of CDK 4/6is to ET in the metastatic setting.¹⁹ Importantly, there was no difference in CBR between the two doses. We did find a numerical 3-month difference in PFS between the two groups, with longer PFS in the 125 mg group; however, this difference was not statistically significant and the small sample size and pretreated population limit efficacy assessment. These data suggest that palbociclib 100 mg may be better tolerated than 125 mg and is a reasonable toxicity management strategy, and also highlights the importance of carefully considering tolerability when establishing initial dosing for new therapies.

Eighteen patients in this trial received a previously administered endocrine partner, with 10 experiencing PFS <6 months, potentially confounding palbociclib dose comparisons. However, eight patients achieved prolonged PFS (11.01-18.66 months) at both the 125 and 100 mg doses, suggesting that the addition of a CDK4/6i may offer clinically meaningful benefit even after disease progression on the same ET.

Although this study focused on palbociclib dose reductions, such adjustments are common across CDK4/6is and may not compromise efficacy in the metastatic setting.²⁰ For ribociclib, ∼60% of patients require dose reductions at 600 mg without apparent impact on efficacy21, 22, 23 and in the AMALEE trial, patients receiving 400 mg had similar ORR to those on 600 mg.²⁴ These results, along with phase III data,^7,8 supported the use of a 400 mg starting dose in the adjuvant NATALEE trial, which showed a significant improvement in invasive disease-free survival with ribociclib and ET compared with ET alone (hazard ratio 0.748, 95% CI 0.618-0.906, P = 0.0014), and a lower rate of grade ≥3 neutropenia (43.8% compared with 60% in trials of 600 mg/day).^6,25

In the metastatic setting, lower yet effective doses of CDK4/6is may enhance tolerability when combined with other potent therapies, particularly in the presence of overlapping toxicities. Several ongoing trials [e.g. NCT04862663, NCT04072952, NCT05563220, TRADE (NCT06001762), CDK study (NCT06377852)] are evaluating lower dose combinations with novel endocrine and targeted agents, highlighting the value of determining the minimal effective biological dose, and increasing support for de-escalation strategies, as proposed by the FDA’s Project Optimus.²⁶

Preclinical and clinical studies have reported that CDK4/6is decrease Rb phosphorylation.^2,10,27,28 Our analysis of tumor and skin tissue before and during treatment showed significant reductions in pRb and Ki-67 at both doses, indicating comparable Rb phosphorylation inhibition and cell cycle arrest. We note that changes in Rb phosphorylation and Ki-67 can be observed in keratinocytes, indicating an appropriate on-target response in normal tissue and consistent with findings of other prospective palbociclib trials assessing the reduction of pRb and Ki-67 in skin.²⁹ These data support the use of pRb and Ki-67 as pharmacodynamic biomarkers for CDK4/6i activity and may serve as a complimentary tool to non-invasive liquid biomarkers of response that are currently under investigation including serum thymidine kinase 1 activity, ctDNA, exosome markers, and micro RNA.^17,30 The comparable pharmacodynamic effects achieved by both dose levels is consistent with the absence of a statistically significant difference in PFS. Although the numerically higher PFS in the 125 mg group in the context of a small sample size indicates further work is required to assess outcomes of these dose levels, the similar biological effects observed suggest that such a comparison is indeed worthwhile.

Our study found that higher baseline Ki-67 in tumor biopsies was associated with shorter PFS, consistent with prior data linking elevated Ki-67 to poorer outcomes in both metastatic and early-stage breast cancer treated with CDK4/6is.^31,32 While no significant associations were observed between PFS and baseline or fold changes in total Rb and pRb, or fold change in Ki-67, this may reflect the small sample size and heavily pretreated cohort. Interpretation of fold changes was further limited by low baseline biomarker levels in some patients. Although baseline pRb may not reliably predict benefit, preclinical and clinical data suggest that Rb loss or degradation contributes to resistance.³³ Notably, PALOMA-3 showed a higher incidence of RB1 loss-of-function mutations in patients treated with palbociclib, highlighting Rb pathway alterations as potential resistance mechanisms.³⁴ It is also possible that pRb and Ki-67 reduction after 3 weeks of exposure is necessary but not predictive of long-term clinical benefit. Some tumors may adapt via alternate cell cycle pathways, while others may enter a senescent state associated with prolonged PFS. Markers predicting a senescent response have been associated with prolonged PFS in studies of CDK4/6 inhibition in liposarcoma.35, 36, 37 Ultimately, a deeper understanding of biological responses following CDK4/6-induced cell cycle arrest may help identify patients most likely to benefit.

Baseline ctDNA analysis revealed TP53 (43%), PIK3CA (39%), and ESR1 (35%) mutations, consistent with prior studies of HR-positive metastatic breast cancer.^34,38,39 No PFS difference was observed between patients with or without ESR1 mutations, possibly reflecting the ability of tamoxifen and fulvestrant to overcome ESR1 mutations.^40,41 PIK3CA and TP53 mutations were associated with poorer PFS, especially when both were present alongside ESR1 mutations, consistent with prior studies.^42,43 However, PFS in patients with PIK3CA tumor mutations is improved with the addition of CDK4/6is to ET and may enhance sensitivity to PI3-kinase inhibitors.44, 45, 46, 47 Conversely, TP53 mutations may confer intrinsic resistance to CDK4/6i by preventing durable cell cycle arrest requiring CDK2 inhibition.^48,49 Interestingly, our analysis showed that in those with mutations in TP53, PIK3CA, and ESR1, there was no correlation between baseline Ki-67 and baseline pRB indicating that cell proliferation may be driven by pRB-independent mechanisms in these patients.

Limitations include the small sample size and a heterogeneous population receiving later-line ET, sometimes after chemotherapy. This trial preceded standard first- and second-line CDK4/6i use for metastatic HR-positive breast cancer. Interestingly, recent data from the Sonia trial⁵⁰ suggest similar survival with first- or second-line CDK inhibition, fueling debate on optimal sequencing. In addition, fewer patients in the 100 mg group received adjuvant chemotherapy or hormone therapy than the 125 mg group, and, as previously mentioned, some patients received endocrine partners previously used. These factors highlight the need for careful evaluation of optimal dosing.

Palbociclib was selected for this trial as the first approved CDK4/6i, with the trial starting shortly after drug approval in the United States. While recent studies suggest variation in the impact on OS among CDK4/6i agents, cross-trial comparisons are limited by variations in study design and patient populations. Notably, real-world data have shown improved median OS with palbociclib plus letrozole versus ET alone, including benefits in patients over 65 years of age.^51,52 Extended follow-up from the PARSIFAL trial reported a median OS of 65.4 months with first-line palbociclib and ET, comparable to outcomes with ribociclib and abemaciclib.⁵³ Additionally, a large real-world analysis of over 9000 patients found no significant OS differences among the three CDK4/6is, supporting a continued role for palbociclib in treating metastatic HR-positive breast cancer.⁵⁴

In conclusion, this randomized phase II trial of palbociclib 125 mg and 100 mg dose in combination with ET demonstrated improved tolerability of the lower dose. Both cohorts demonstrated similar decreases in Rb phosphorylation and in Ki-67 expression, indicating similar and robust molecular responses. These data provide correlative as well as clinical data further confirming that appropriate dose reductions can improve tolerability with similar on-target effects. It also reinforces the feasibility and need for biomarker-driven approaches to dose optimization, supporting the approach of the FDA’s Project Optimus, as well as the importance of ctDNA analyses to extend findings to patients harboring deleterious mutations. Finally, this work illustrates the complexity of dose optimization to achieve efficacy with maximal tolerability.