Inclusion Criteria

Inclusion Criteria

- Patients with histologic proof of adenocarcinoma of the esophagus (including GE
junction) or adenocarcinoma of the stomach will be eligible.

- Location of the tumor for assignment to study will be determined by prior endoscopy
or barium contrast study.

- Patients must have either metastatic or recurrent cancer.

- Patients must have bidimensionally measurable disease, as defined in Section 8.4.1.1,
page 16. Mediastinal or hilar lymph nodes assessed by CT or MRI scans must be at
least 2 cm in the largest dimension to be considered measurable.

- Prior limited radiation therapy is permitted. Prior radiotherapy must not have
included major bone marrow containing areas (pelvis, lumbar spine), or contained the
single evaluable lesion in a radiation field. A recovery period of at least 4 weeks
after completion of radiotherapy is required prior to study treatment.

- Patients must have an anticipated life expectancy of at least 12 weeks.

- Patients must have a performance status of 0 or 1 on the ECOG performance scale.

- Patients must be > 18 years old.

- Patients must give written informed consent prior to enrollment.

- Patients should have adequate organ function defined as follows: Absolute
granulocytes > 1,500/mm3 and platelets > 100,000/mm3; serum bilirubin < 1.5 mg/dL and
SGOT < 3X upper limit of institutional norm; and serum creatinine < 1.5 mg/dL.

- Patients must have recovered from recent surgery. One week must have elapsed from
the time of a minor surgery and 3 weeks from major surgery.

Exclusion Criteria

- Patients who have received more than one prior chemotherapy regimen or immunotherapy
for metastatic disease. Prior 5-FU alone as an adjuvant therapy or radiosensitizer
is not counted. Patients must have an interval of 4 weeks from prior chemotherapy or
immunotherapy with full recovery.

- Patients receiving concurrent chemotherapy, immunotherapy, or radiotherapy.

- Patients who are potentially curable with a chemotherapy, radiotherapy, surgery, or
any combination of the above.

- Patients with known brain metastases.

- Patients with a history of seizures or are receiving phenytoin, phenobarbital, or
other antiepileptic prophylaxis.

- Pregnant or lactating women. All women of childbearing potential must have a
negative pregnancy test prior to entry into the study. All patients of procreative
potential must be advised of the importance of avoiding pregnancy and using
appropriate methods of contraception while participating in this investigational
trial.

- Patients with serious intercurrent infections, or any other concurrent disease which,
in the investigator's judgment, would make the patient inappropriate for entry into
this study.

- Patients with psychiatric disorders rendering them incapable of complying with the
requirements of the protocol.

- Patients with osseous metastasis as only site of disease.

- Patients with any concurrent active malignancy other than non-melanoma skin cancers
or carcinoma-in-situ of the cervix. Patients with previous malignancies but without
evidence of disease for > 5 years will be allowed to enter the trial.

- Patients with known Gilbert's syndrome.

- Patients who have a significant clinical neuropathy of greater than grade 2.

- Patients with unstable angina, a history of myocardial infarction within the previous
6 months, or current clinical evidence of congestive heart failure. Patients taking
medication for congestive heart failure and showing no clinical signs or symptoms are
eligible.

- Patients who have previously received a taxane or campthothecin

- Patients who have received any investigational therapy within the previous 4 weeks.