Because the role of chemotherapy, interferon, or somatostatin analogs as antiproliferative agents is uncertain, currently few treatment options exist for patients with metastatic or inoperable gastroenteropancreatic neuroendocrine tumors (GEP-NET). Fifty-eight patients with somatostatin receptor-positive GEP-NET were treated in a phase I dose-escalating study with cumulative doses of 47 mCi to 886 mCi of the radiolabeled somatostatin analog [(90)Y-DOTA(0),Tyr(3)]-octreotide. At baseline, 47 patients had progressive disease, and 36 were symptomatic. The extent of disease was: 4 patients without liver metastases and 52 patients with liver metastases, including 16 patients with very advanced disease, qualified as "end-stage," and 2 end-stage patients without liver metastases. The objective responses were 5 partial response (PR), 7 minor response (MR), 29 stable disease (SD), and 17 PD. Overall, 33 patients (57%) experienced some improvement in their disease status, including conversion from PD into SD and improvement from SD into MR. Accordingly, 21 of 36 patients (58%) had improvement in Karnofsky performance score or symptoms. The median overall survival (OS) was 36.7 months (95% confidence interval [CI] 19.4-54.1 months). The median progression-free survival in 41 patients who had at least stable disease at the end of the treatment period was 29.3 months (95% CI 19.3-39.3 months). Patients who had SD at baseline had a significantly better OS than patients who had PD at baseline. The extent of disease at baseline also was a significant predictive factor for OS. The OS after therapy with [(90)Y-DOTA(0),Tyr(3)]-octreotide was significantly better than in a historic control group of 32 comparable patients with GEP-NET who had been treated with another radiolabeled somatostatin analog, [(111)In-DTPA(0)]-octreotide (median OS 12.0 months, 95% CI 6.2-17.8 months). The difference in OS for both therapies remained highly significant in a multivariate Cox proportional hazard model including progression status and extent of disease at baseline as covariates. Although the objective response after therapy with [(90)Y-DOTA(0),Tyr(3)]-octreotide by standard criteria seems modest, the significantly longer OS compared with historic controls is most encouraging.