ABSTRACT: VEGF blockade does not uniformly result in clinical benefit. We evaluated safety, dose-limiting toxicities (DLT), recommended phase II dose (RP2D), antitumor efficacy, and exploratory biomarkers including pharmacogenomics and pharmacokinetics with sorafenib, bevacizumab, and paclitaxel in patients with refractory cancers. The study had a "3 + 3" design, using paclitaxel 80 mg/m2 every week for 3 weeks, in every 4 week cycles, bevacizumab 5 mg/kg every 2 weeks, and sorafenib 200 or 400 mg twice a day, 5 or 7 days/week (5/7, 7/7). The MTD cohort was expanded. Twenty-seven patients enrolled in 3 cohorts: sorafenib 200 mg twice a day 5/7, 200 mg twice a day 7/7, and 400 mg twice a day 5/7. DLTs were grade 3 neutropenia >7 days (cohort 1, 1), grade 3 hypertension (cohort 2, 1), grade 3 hand-foot skin reaction (HFSR; cohort 3, 2). MTD was sorafenib 200 mg twice a day 7/7. Six DLTs occurred in cohort 2 expansion: grade 3 HFSR (2), grade 2 HFSR with sorafenib delay >7 days (2), grade 4 cerebrovascular accident (1), grade 3 neutropenia >7 days (1). RP2D was sorafenib 200 mg twice a day 5/7. Most patients (62%) dose reduced sorafenib to 200 mg daily 5/7 after a median 3 (range, 2-17) cycles. Response rates were 48% overall (27) and 64% for ovarian cancers (14). VEGF-A-1154AA and -7TT recessive homozygous genotypes conferred worse overall survival versus alternative genotypes (7 vs. 22 months). Intermittent, low-dose sorafenib (200 mg twice a day 5/7) combined with bevacizumab and paclitaxel was tolerable and had high antitumor efficacy in patients with refractory cancer (NCT00572078).