Gemcitabine (G) and docetaxel (D) are commonly used to treat recurrent/metastatic soft tissue sarcoma. This study tested the hypothesis that outcomes would be improved by addition of bevacizumab (B). The initial design was randomized double-blind trial of G + D + B versus G + D + placebo. Due to slow accrual this was modified to single-arm open-label G + D + B. Eligible patients had diagnosis of leiomyosarcoma, pleomorphic undifferentiated sarcoma, pleomorphic liposarcoma, or angiosarcoma. Treatment was B 15 mg/kg on d1, G 900 mg/m(2) on d1 and d8, and D 75 mg/m(2) on d8, q21d. Primary endpoint was progression-free survival (PFS) at 6 months and would be met if ≥17 patients were progression-free at 6 m. Secondary endpoints are response rate, PFS at 3 m, overall survival, and toxicity. Of 44 patients enrolled, 35 were treated with GDB and evaluable for safety and efficacy. Median age was 55, 50% male, most ECOG 0. Toxicity is mostly myelosuppression with one deep vein thrombosis and one small bowel perforation possibly related to B. There were 17 partial responses (49%) by RECIST 1.1. Among 35 patients, the number who remained on study and progression-free was 24 at 3 m and 15 at 6 m. 9 withdrew prior to 6 m for reasons other than toxicity or progression. PFS at 6 m was 65% (95% CI: 51-85%). The primary endpoint of 6 m PFS was not met due to censoring of patients who withdrew. However PFS at 3 m (76%) was promising and response rate was higher than expected from G + D.