Patients with metastatic or recurrent gastric adenocarcinomas, who had failed (primary or secondary after initial response) 5-FU-based chemotherapy within 1 year before their registration, were enrolled in this study. The patients were required to have a measurable disease, a life expectancy of 8 weeks or more, an ECOG performance status of 2 or less, adequate bone marrow (neutrophil count >1,000/mm³, and a platelet count >100,000/ mm³), hepatic (serum total bilirubin level <1.5 mg/dl, AST/ALT <3X ULN) and renal function (serum creatinine level <1.5 mg/dl). This trial was designed to detect a response rate of 30% compared with a minimal, clinically meaningful response rate of 10%, with a statistical power of 80% for accepting the hypothesis and 5% significance for rejecting the hypothesis. The total sample size required to allow for a follow-up loss rate of up to 10% was 35 patients with a measurable disease; all patients gave their written informed consent.

The treatment consisted of paclitaxel, 175 mg/m², in a 3-hour infusion followed by cisplatin, 60 mg/m², in a 1-hour infusion. Each cycle of chemotherapy was given every 3 weeks if the patient's blood count had returned to normal and non-hematological toxic effects had resolved. The dosage of the subsequent cycle was adjusted according to the toxic effects that had developed during the preceding cycle. All patients received a standard supportive regimen, consisting of hydration with normal saline at least 3 L/24 hours, dexamethasone and 5-HT3 inhibitors. Follow-up history and physical examinations, tumor measurements and toxicity assessments were performed before each 3-week cycle of therapy.

The tumor response was evaluated according to the Response Evaluation Criteria in Solid Tumors (RECIST) criteria (14), and was assessed by a computed tomographic scan of the abdomen and pelvis, and by the same tests initially used to stage the tumor. All tumor measurements were recorded by radiologists, in millimeters, for all measurable lesions. Responses were assessed after 2 and 4 courses of chemotherapy. Toxicity grading was based on the National Cancer Institute common toxicity criteria (NCI-CTC version 3.0). Patients were evaluated for toxic effects before each new treatment cycle. The survival duration was computed as the time between the first day of the second-line chemotherapy given as the protocol and the patient's death or, in the case of continuing survival, the last day of follow-up. The overall and progression-free survival curves were computed using the Kaplan-Meier method.

A total of 37 patients were registered between Feb. 2002 and Sep. 2006. The characteristics of the eligible patients are listed in Table 1. The male to female ratio was 21 : 16. The median patient age was 56 years (range: 28~79), and 73% of the patients had an Eastern Cooperative Oncology Group performance status of 0 or 1. Prior chemotherapy consisted of 5-fluorouracil- based or cisplatin-based regimens in almost all patients. Fifteen patients (40%) had previously undergone chemotherapy with more than 2 regimens. Eighteen patients had at least two organs involved. The majority of the patients had metastatic lymph nodes (51%), and 30% had liver metastases.

Of the 37 patients assessable for survival, 5 were not "evaluable" for response. Three patients had refused treatment prior to evaluation and 2 were lost to follow-up. Thus, 32 patients were assessable for response. 4 partial responses were obtained, corresponding to 13% of the evaluable patients. There were 13 patients with a stable disease (40%) and 15 progressed (47%) (Table 2). The median follow-up duration was 37 months. The median progression-free survival was 4 months (95% CI : 2~6 months) (Fig. 1). The median overall survival was 12.6 months (95% CI : 5.5~19.7 months), with a 1-year survival rate of 54% (Fig. 2).

Thirty-seven patients received a total of 135 treatment cycles, and received a median of 3 courses of paclitaxel plus cisplatin (range, 1~9 courses). The planned dose intensities of paclitaxel and cisplatin were 58.3 and 20 mg/m²/week respectively. The relative dose intensities (the actual dose delivered as a proportion of the planned dose) of paclitaxel and cisplatin were 88 and 93%, respectively. The toxicity was evaluable in all 37 patients and for all 135 chemotherapy cycles. The most frequent hematological toxicity was anemia (grade 1 or 2 in 52% of treatment cycles). Grades 3 or 4 neutropenia were observed in 18 cycles (14%), but only 6 episodes of nonfatal neutropenic fever were observed. Grade I or II nausea/vomiting was observed in 7 patients (19%), and grade II neuropathy was reported in 5 patients (14%). One patient experienced an anaphylactic reaction, but there were no treatment-related deaths (Table 3).