PATIENTS AND METHODS

RESULTS

From June 1987 through October 1996, 44 member institutions of the GOG and the North Central Cancer Treatment Group (NCCTG) enrolled 267 patients in this study. Of the 65 patients found to be ineligible, 10 patients did not have FIGO stage III and six had tumors of low malignant potential. Other pathologic exclusions were as follows: ineligible primary (5), second primary (1), and inadequate pathology material to review (1). The SLL operative procedure did not conform to protocol requirements in 41 patients, and one additional patient had a positive SLL. The most common reason for this was the lack of bilateral para-aortic lymph node sampling. The remaining 202 eligible patients constitute the basis of this report.

There were 104 eligible patients randomly selected to receive IP ³²P and 98 patients assigned to NFT. The distribution of patient characteristics by regimen is displayed in Table 1⇓. All but one patient had received prior chemotherapy with cisplatin or a cisplatin analog, and 43 patients (44%) in the NFT group and 36 patients (35%) in the ³²P group had additionally received paclitaxel. The prevalence of liver surface involvement at the time of the initial debulking surgery among the IP ³²P group was approximately twice that of the NFT group; this was the only noteworthy imbalance. The median number of courses of cisplatin administered for first-line therapy was six for both arms (range, 0 to 10 courses for ³²P, and 3 to 12 courses for the NFT regimen).

Treatment data for patients who were randomly selected to receive IP ³²P are presented in Table 2⇓. Sixteen patients (15%) did not receive IP ³²P, primarily due to inadequate distribution of ^99mTc in the peritoneal cavity.⁹ Five did not receive therapy based on medical reasons, including inability to place catheter, and the remaining two patients refused treatment for nonmedical reasons.

The adverse effects data are presented in Table 3⇓. There was a statistically significant higher frequency of grade 1 to 2 gastrointestinal toxicity, with 16 (15%) patients on the IP ³²P arm and five (5%) patients on the control arm experiencing that effect. The two grade 3 to 4 gastrointestinal toxicities in the control regimen include one patient with a fistula and one with elevated AST and ALT (liver laceration). There were three bowel obstructions requiring surgery among patients in the ³²P group. There were no other noteworthy differences in reported toxicities.

There were 131 recurrences, with 63 (64%) on the NFT regimen and 68 (65%) on the IP ³²P regimen. These recurrences, plus 10 deaths without recurrence (five in each arm) contributed to the RFS curves displayed in Figure 1⇓. The 5-year RFS rate for the IP ³²P and NFT groups were 42% and 36%, respectively. The relative risk of recurrence is 0.90 (IP ³²P to NFT group; 90% confidence interval [CI], 0.68 to 1.19).

There was no statistically significant difference in RFS between the two regimens (log-rank test, P = .27, one-tailed). The median RFS was 32.9 months for the NFT and 43.3 months for the IP ³²P arm. The recurrence rate over time did not decrease, which suggests that the number of recurrences will continue to increase with further follow-up. There were 66 patients (33%) with peritoneal and 52 patients (26%) with extraperitoneal recurrence. There is no difference between arms regarding distribution of the first reported site of recurrence (Table 4⇓).

At the time of the analysis, there were 111 patients (55%) who died of their cancer. Ten patients died of other causes, and the cause of death in one patient is unknown. Median follow-up for living patients was 63 months. There was no significant difference (log-rank test, P = .19, one-tailed) in OS between the two treatment groups (Fig 2⇓). The relative risk of death was 0.85 (IP ³²P to NFT group) (90% CI, 0.62 to 1.16). The 5-year OS rate was 67% and 63% for IP ³²P and NFT groups, respectively. The following factors were analyzed for their relationship to RFS and OS: age, performance status, cell type (clear-cell; endometrioid; mixed epithelial; serous; adenocarcinoma; unspecified; and all others), histologic tumor grade, site(s) of involvement, residual disease (microscopic, gross but optimal, and suboptimal) at initial surgery, type and amount of first-line chemotherapy. There was no difference in RFS for patients who received prior paclitaxel compared to those who did not. The analysis showed initial disease in the omentum, and gross residual disease at the completion of the initial surgery, to be the only statistically significant prognostic factors for both RFS and OS. Age at diagnosis was borderline significant (P = .07) for OS only. From the modeling of RFS, gross disease was associated with a 36% increase in risk and omental disease was associated with 85% increased risk. Exploratory analysis of the six distinct cell types revealed that only the other cell types (undifferentiated [eight], transitional cell [three], and villoglandular [one]) were associated with recurrence and a more than two-fold increase in the risk of death when compared with serous adenocarcinoma.

DISCUSSION

In this trial, IP ³²P did not decrease the risk of relapse or improve survival for patients with stage III epithelial ovarian cancer after a negative SLL. Despite complete clinical remission after initial surgery and platinum-based chemotherapy and no pathologic findings of persistent disease at SLL, 61% of stage III ovarian cancer patients had tumor recurrence within 5 years of negative SLL.

The use of melphalan after cytoreductive surgery for ovarian cancer was followed by SLL to assess the surgical-pathologic response and reduce toxicity associated with prolonged chemotherapy.^1–,⁵ Among patients with complete clinical response, only 50% were found to have negative SLL. However, negative SLL had raised the possibility of cure, and further treatment was discontinued.

It was recognized that complete surgical responses, more commonly achieved with cisplatin than with melphalan therapy, were not durable.^8–,¹¹ Two large studies confirmed these disappointing data.^10,11 Rubin et al¹⁰ reported 54% of 63 stage III or IV patients had tumor recurrence after initial platinum-based chemotherapy and negative SLL with a median follow-up of 55 months. Bolis et al¹¹ reported 44% recurrences in 140 stage III or IV patients treated with a platinum regimen and who had a negative SLL.

The most common sites of recurrence after negative SLL were the pelvis and upper abdomen as reported by Gershenson et al.⁹ Rubin et al¹⁰ noted that 60% of recurrences were intraperitoneal. These and other studies demonstrated that subclinical residual disease remained despite negative findings at SLL. This patient population represented an opportunity for consolidation therapy geographically focused on the peritoneal cavity. The rationale for the use of ³²P in this study was based on the high peritoneal recurrence rate from subclinical disease and the radiation characteristics of ³²P.

Radioactive ³²P, with a physical half-life of 14.3 days, decays by pure beta emission with mean energy of 695 keV. The biologic effect of the isotope in colloid suspension remains within 4 mm, thus primarily treating the peritoneal surface. The absence of gamma decay limits the radiation dose to other intra-abdominal organs such as the liver, kidneys, and intestines and permits safe treatment of large peritoneal surfaces. The ³²P colloid preparation is not absorbed into the systemic circulation and has no hematologic toxicity associated with other ³²P compounds.

Hester and White,²⁸ in one of the earliest reports on the use of ³²P in gynecology, noted that ³²P was used for malignant effusions and that ovarian cancer patients often had positive washings or cytology. Pezner et al²⁹ treated 91 patients with stage I or II disease that was totally resected, or with ≤ 8 mm residual disease. The investigators concluded that ³²P might be of value as adjuvant therapy for patients with stage I or II disease.

Additional reviews were published in the late 1970s,^13,14 and by the end of the 1980s several authors evaluated the efficacy of ³²P after initial operation for stage I or II ovarian cancer patients.^15,16 In 1990, Young et al¹⁵ published the results of the GOG/OCSG randomized clinical trial in which 141 stage I or II patients underwent surgery followed by either 15 mCi of ³²P or oral melphalan. The RFS was similar between the two regimens and there was no difference in patterns of recurrence. IP ³²P appeared to have the same efficacy as melphalan without the risk of leukemia.

Also appearing in the literature are several reports of nonrandomized studies using ³²P after SLL (Table 5⇓).^12,17–,¹⁹ Spencer et al,¹⁹ in a nonrandomized study, reported tumor in 4 of 17 patients with no further treatment after negative SLL, but none in 14 patients treated with IP ³²P. Varia et al¹² at the University of North Carolina reported a large retrospective series in which IP ³²P was administered on the same day or immediately after grossly negative SLL to ensure free distribution of ³²P and before the microscopic pathology report was available. In that study, 118 patients underwent SLL and, of the 57 patients with negative SLL, 43 received 15 mCi IP ³²P. The 4-year survival in the negative SLL group was 89% with IP ³²P as compared with 67% for those not receiving ³²P due to adhesions or other study parameters. In the microscopic/minimal disease group, the 4-year survivals were 59% and 22%, with and without IP ³²P, respectively. Bowel toxicity was low and comparable in both groups. Although there was no randomization in this trial and patients with stage I to III disease were included, 59% 4-year survival in the microscopic/minimal disease group receiving IP ³²P suggested a therapeutic role for IP ³²P.

Two reports with overlapping patient populations presented a less optimistic view of the value of ³²P.^30,31 In the first report, Vergote et al³⁰ presented data on 340 patients who were randomly selected to receive cisplatin (171 patients) or ³²P (169 patients) as adjuvant therapy (169 patients). Patients had stage I–III ovarian cancer with no residual disease after surgery. Twenty-eight patients randomly selected to ³²P did not receive ³²P on account of adhesions and received whole abdominal radiation (WAR) therapy. Intraperitoneal distribution assessment of the colloid was not performed. After a median follow-up of 62 months, the disease-free interval (DFI) was similar (81% v 83%) in both groups at 5 years. Grade 2 bowel obstruction was more common in the ³²P-treated (± WAR) group. Nine patients with small bowel obstruction required surgery: six (9%) in the ³²P-only group and three (21%) in the WAR group. Although survival outcomes were similar, the authors concluded that cisplatin should be used as standard adjuvant treatment because of the morbidity observed with ³²P.

The second report by these investigators included 245 patients with primary disease and 68 with negative SLL.³¹ In this study, ³²P was administered within 6 weeks of operation using doses of 7, 8.1, or 10 mCi. Scintigram was performed to assess distribution of the isotope after administration of ³²P. Bowel obstruction occurred in three (4%) of the 68 patients in the SLL group. The authors concluded with a recommendation for a randomized clinical trial of ³²P versus no further therapy in the negative SLL population.

Gastrointestinal toxicity became an increasing focus in several of these trials,^15,19,32 particularly when external radiation was administered.^29,31,33 Spanos et al³⁴ proposed that the timing of colloid administration was more relevant than the use of external radiation or the number of prior operations.

The results of the randomized clinical trial presented in this report show no statistically significant difference in RFS or OS. The relative risk (experimental to control) was 0.90 (90% CI, 0.68 to 1.19) for recurrence and 0.85 (90% CI, 0.62 to 1.16) for survival when adjusting for gross residual disease status at initial surgery. There may be several explanations for the lack of beneficial effect. First, there is the possibility that the therapy itself is not effective. Second, the beneficial effect of radiation is limited to a depth of 3 to 5 mm and cannot adequately treat epithelial ovarian cancer cells to a greater depth. This would be particularly true for retroperitoneal nodes and parenchymal liver or other abdominal organs. Third, ^99mTc distribution scan and the actual distribution of IP ³²P may not have reached all of the target peritoneal surfaces and volumes at risk. Fourth, the volume of occult tumor burden despite a negative SLL is too great for IP ³²P alone in stage III patients.

At least 50% of optimal stage III ovarian cancer patients in clinical remission after paclitaxel and platinum-based contemporary chemotherapy have overt disease at SLL.³⁵ In addition, 50% or more with negative SLL have undetected occult disease that manifests as recurrent disease within 2 years. This signifies that, at a minimum, 75% of patients have persistent disease despite apparent complete clinical remission. Improvements in first-line therapy with or without new consolidation strategies are clearly needed to reduce the incidence of persistent disease.

APPENDIX

The following GOG institutions participated in the study: University of Alabama at Birmingham, Birmingham, AL (CA 12484); Duke University Medical Center, Durham, NC (CA 12534); Abington Memorial Hospital, Abington, PA; University of Rochester Medical Center, Rochester, NY; (CA 12482), Walter Reed Army Medical Center, Washington, DC (CA 23501); University of Minnesota Medical School, Minneapolis, MN (CA 23088); University of Mississippi Medical Center, Jackson, MS (CA 13633); Colorado Gynecologic Oncology Group, P.C., Denver, CO (CA 15975); University of California Medical Center at Los Angeles, Los Angeles, CA (CA 13630); Milton S. Hershey Medical Center, Hershey, PA (CA 16386); Georgetown University Hospital, Washington, DC (CA 16938); University of Cincinnati, Cincinnati, OH; University of Iowa Hospitals and Clinics, Iowa City, IA (CA 19502); Indiana University Medical Center, Indianapolis, IN (CA 21720); Wake Forest University School of Medicine, Winston-Salem, NC (CA 21946); Albany Medical College, Albany, NY (CA 27469); University of California Medical Center at Irvine, Irvine, CA (CA 23765); Tufts-New England Medical Center, Boston, MA (CA 37569); Rush-Presbyterian-St. Luke’s Medical Center, Chicago, IL (CA 12485); University of Kentucky, Lexington, KY; The Cleveland Clinic Foundation, Cleveland, OH; Johns Hopkins Oncology Center, Baltimore, MD; Eastern Pennsylvania Gynecology/Oncology Center, P.C., Philadelphia, PA; Washington University School of Medicine, St. Louis, MO; Cooper Hospital/University Medical Center, Camden, NJ; Columbus Cancer Council, Columbus, OH; North Central Cancer Treatment Group, Mayo Clinic, Rochester, MN; M.D. Anderson Cancer Center, Houston, TX; University of Massachusetts Medical Center, Worcester, MA; Fox Chase Cancer Center, Philadelphia, PA; Medicial University of South Carolina, Charleston, SC; Women’s Cancer Center, University of Oklahoma Health Science Center, Oklahoma City, OK; University of Chicago, Chicago, IIL; Tacoma General Hospital, Tacoma, WA; Case Western Reserve University, Cleveland, OH; Tampa Bay Cancer Consortium, Tampa, FL; Brookview Research Inc., Nashville, TN