A prospective randomized trial using the group sequential statistical method with frequent planned interim analyses was initiated in patients with metastatic colorectal adenocarcinoma comparing methotrexate and 5 fluorouracil (MTX-FU) to methyl CCNU, vincristine, FU, and streptozotocin (MOF-Strep). As a requirement of the group sequential design, a p value of 0.0125 was necessary to stop the trial at the first analysis. The first analysis was done after 17 patients were entered in each study arm and revealed six partial responses to MOF-Strep (35%) and one partial response to MTX-FU (6%) (p = 0.017). As a result, patient accrual continued. Problems encountered with the implementation of the group sequential methodology and the importance of stratification for prognostic variables are discussed.