Third, although the efficacy of initial surgical cytoreduction in patients with stage III disease has been accepted on the basis of retrospective analyses, prospective trials are needed to address several important questions. One such study suggests an advantage for interval surgical cytoreduction at the midpoint of a series of chemotherapy courses. Prospective randomized trials of initial or secondary surgical cytoreduction have not been completed. Investigations of the relative merits of each of these approaches versus no surgery are needed, as well as trials evaluating which of these points in the therapy represents the optimal time to introduce surgical resection into the management of advanced disease. Such studies are difficult to conduct because of the widespread acceptance of the role of surgical cytoreduction in ovarian carcinoma. A confirmatory trial of interval cytoreduction is nearing completion, but further study is needed.

Fourth, efforts continue to investigate the role of new agents in the management of ovarian carcinoma. Current interest continues to center on further delineation of the role of paclitaxel. The plethora of new agents with activity in patients who are clinically resistant to the platinum compounds and paclitaxel, however, opens the possibility for the addition of clinically non-cross-resistant drugs to front-line paclitaxel-platinum therapy. Defining the role of these new agents is of paramount importance.

Finally, dose intensity continues to command significant interest. Three basic ways to enhance the dose intensity have been proffered: escalation of dose within the range that can be achieved without marrow reconstitution, high-dose chemotherapy with support of autologous bone marrow transplant or peripheral stem cell transfusion, and, in the case of ovarian carcinoma, intraperitoneal administration of drug. Eight randomized trials of dose escalation over a standard range of doses have been completed. Six show no advantage to a doubling of dose intensity, and the other two have major design problems. Further exploration of this approach seems unwarranted.