Introduction: Phosphatidylinositol 3-kinase (PI3K) pathway activation in breast cancer (BC) is associated with tumor growth and resistance to anticancer therapies, including paclitaxel (PAC). Combined treatment with the pan-Class I PI3K inhibitor buparlisib (BKM120) + PAC has shown signs of activity in a Phase Ib study in advanced solid tumors, and may delay chemoresistance and progression on PAC therapy in BC.

Study design: BELLE-4 (NCT01572727) is a randomized, double-blind, placebo-controlled, adaptive, Phase II/III seamless study of buparlisib + PAC in women with HER2- locally advanced or metastatic BC with no prior systemic chemotherapy for advanced BC (ABC).

Patients were randomized (1:1) to receive buparlisib (100 mg/day) or placebo with PAC (80 mg/m2/week). Randomization was stratified by PI3K pathway status (activated vs non-activated) and hormone receptor status.

The primary endpoint was progression-free survival (PFS; locally assessed per RECIST v1.1) in the full and PI3K pathway-activated populations. Secondary endpoints included overall survival, overall response rate (ORR), clinical benefit rate, and safety (CTCAE v4.03).

An adaptive interim analysis was planned at the end of Phase II after ≥125 PFS events had occurred. Decision rules based on predefined thresholds of PFS predictive probabilities of success in the full and PI3K pathway-activated populations were used to decide whether the study would enter Phase III (in the full population or restricted to the PI3K pathway-activated population), or be stopped for futility.

Results: As of June 7, 2014, 338 women were randomized to receive buparlisib + PAC (n = 168) or placebo + PAC (n = 170). In total, 125 (37%) patients had PI3K pathway-activated tumors, and 248 (73%) tumors were hormone receptor-positive.

At data cut-off, 112 (67%) vs 93 (55%) patients (buparlisib vs placebo arm) had discontinued treatment, most commonly for disease progression (30% vs 35%) and adverse events (AEs; 22% vs 8%). Median duration of study treatment exposure was 3.4 months (range: 0.2-16.2) vs 4.6 months (range: 0.2-20.1). The most frequent AEs (≥35% of patients) were diarrhea (55% vs 34%), alopecia (49% vs 52%), nausea (45% vs 27%), hyperglycemia (45% vs 11%), rash (40% vs 24%), fatigue (37% vs 34%), and neutropenia (35% vs 30%). Grade 3/4 AEs occurred in 79% vs 52% of patients.

At the interim analysis (based on 131 PFS events), the study met the futility criteria in the full and PI3K pathway-activated populations as per the predefined decision rules. Primary efficacy results (buparlisib vs placebo arm) are as follows: In the full population, hazard ratio (HR) was 1.18 (95% CI: 0.82-1.68) with median PFS 8.0 vs 9.2 months. In the PI3K pathway-activated population, HR was 1.17 (95% CI: 0.63-2.17) with median PFS 9.1 vs 9.2 months. In the PI3K pathway non-activated population, HR was 1.18 (95% CI: 0.76-1.83) with median PFS 7.3 vs 9.2 months. ORR in the full population was 23% vs 27%, with similar trends in the PI3K pathway-activated and non-activated populations.

Conclusions: At the interim analysis, the predefined futility criteria were met and Phase III was not initiated. No PFS benefit was conferred by addition of buparlisib to PAC in either the full or PI3K pathway-activated populations, possibly due to limited treatment exposure; the PI3K pathway may not drive PAC resistance in untreated HER2- ABC.

Citation Format: Miguel Martín, Arlene Chan, Luc Dirix, Joyce O'Shaughnessy, Roberto Hegg, Alexey Manikhas, Mikhail Shtivelband, Petr Krivorotko, Norberto Batista López, Mario Campone, Manuel Ruiz Borrego, Qamar J. Khan, J. Thaddeus Beck, Manuel Ramos Vasquez, Patrick Urban, Pantelia Roussou, Emmanuelle Di Tomaso, Cristian Massacesi, Suzette Delaloge. BELLE-4: A Phase II study of buparlisib and paclitaxel in women with HER2- advanced breast cancer, with or without PI3K pathway activation. [abstract]. In: Proceedings of the AACR-NCI-EORTC International Conference: Molecular Targets and Cancer Therapeutics; 2015 Nov 5-9; Boston, MA. Philadelphia (PA): AACR; Mol Cancer Ther 2015;14(12 Suppl 2):Abstract nr A166.