Epithelial ovarian cancer is the fifth most frequent cause of cancer death in women and remains the leading cause of gynecologic cancer–related deaths in the United States and Europe.1,2 Standard treatment of advanced ovarian cancer is primary surgery aiming at complete resection followed by platinum and paclitaxel chemotherapy.3

Approximately 15% of epithelial ovarian cancer cases are diagnosed as International Federation of Gynecology and Obstetrics (FIGO) stage IV disease.4 Overall, median survival for patients with stage IV disease is approximately 15–23 months, with an estimated 5-year survival rate of only 20%5–9 Therefore, it is necessary to improve standard treatment, including both surgical management and chemotherapy.

Postoperative residual tumor has been described to be one of the most important prognostic factors in advanced ovarian cancer10–12 In addition, surgical outcome is one of the few prognostic factors that can be influenced by treatment. Performance and outcome of surgical treatment depend on two factors: resectability of the tumor and patient’s condition to tolerate extensive surgical procedures. Other influencing criteria are surgical skill, infrastructure, and capacity.13,14 However, patients with complete resection may have cancers that show a different biology compared with those that are deemed inoperable.

Previous studies of prognostic factors in stage IV ovarian cancer have focused on surgical cytoreduction and the association between residual tumor size and clinical outcome5–7,15–18 However, interpretation of these studies is difficult because of the heterogenous and retrospective study design. There are many studies that use different definitions of “optimal residual tumor” (e.g., <0.5 cm; 1 cm; 2, 3, or 5 cm).15 In addition, many trials included patients receiving chemotherapy other than the platinum- and paclitaxel-based regimens that are considered to be a contemporary standard.

The present exploratory analysis is to our knowledge the largest analysis of patients with FIGO stage IV ovarian cancer treated homogeneously with platinum- and paclitaxel-based chemotherapy within three prospective, randomized phase III trials of the AGO-OVAR (OVAR-3/-5/-7). The aim of this study was to evaluate prognostic factors and the clinical impact of residual tumor and to define the optimal residual tumor after surgical cytoreduction in FIGO stage IV disease.

Patients and Methods

Patients and Study Design

The data set was retrieved from three prospective, randomized phase III trials for advanced epithelial ovarian cancer coordinated and performed by the AGO-OVAR together with GINECO (Groupe d'Investigateurs Nationaux pours l'Étitude des Cancers Ovariens et du sein), a French multicenter study group. All trials were fully published, and details were reported elsewhere.19–22 The inclusion and exclusion criteria were similar in all three trials, and after providing written consent, patients were enrolled onto the study if they had previously untreated epithelial ovarian cancer of FIGO stage IIB–IV.4,23 A total of 573 of 3388 patients had FIGO stage IV disease. Patients were randomly assigned to one of two treatment arms consisting of either carboplatin or cisplatin and paclitaxel, or a combination of carboplatin and paclitaxel versus the same combination with epirubicin or topotecan added in the experimental arms. All patients were scheduled to receive at least six courses of platinum-taxane intravenously every 3 weeks. None of the trials showed any marked difference for the various treatment arms with respect to progression-free survival (PFS) or overall survival (OS). Furthermore, PFS and OS did not depend on the trial the patient belonged to. Therefore, we combined the data sets for this analysis.

Some of the originally enrolled patients were excluded from this analysis. Mainly, these patients had been defined as ineligible in the original study reports. In addition, we excluded patients from this exploratory analysis if metric data about intra- and postoperative findings or details about surgical procedures were not completely documented.

Subgroups based on the localization of FIGO stage IV disease included patients with malignant pleural effusions, parenchymal liver metastases, or any other site of distant metastases.

The measurements of residual tumor diameter were provided by the primary surgeon and were documented postoperatively on the original case report form. We used this commonly reported classification of residual tumor sizes (complete resection and no visible residuals postoperatively, vs. patients with small postoperative residuals and largest diameter of 1–10 mm, vs. patients with residual tumor exceeding 10 mm in largest diameter) both for comparability and adherence to the stratification rules in the original study protocols. Furthermore, we performed a pretest to determine whether any other cutoffs would produce a more marked separation of prognostic subgroups. For this purpose, we analyzed the effect of residual tumor size as a continuous variable. This analysis did not suggest other cutoffs (data not shown).

Recommended standard surgery included median laparotomy, bilateral oophorectomy, hysterectomy, omentectomy, radical tumor debulking including peritoneal stripping, and bowel surgery if indicated. In case of optimal tumor reduction, pelvic and para-aortic lymphadenectomy were also recommended. However, the extent of surgery was finally determined by the gynecologic surgeon, and no patient was excluded on the basis of the chosen procedure.

Postoperative chemotherapy was assigned by central randomization stratified by center and stratum. Stratum 1 included patients with FIGO stage IIB–III disease and postoperative residual tumor size of 0 to 1 cm; stratum 2 included patients with either FIGO stage IV disease or residual tumor sized >1 cm. OS and PFS were calculated from the day of randomization in the original trial. PFS and OS after diagnosis of recurrence were calculated from the day the first recurrence, progression, or death without prior diagnosis if recurrence was observed. Survival data of the primary study collective (stage IV) were compared with stage IIIC patients.

Statistics

OS was determined from the time of random assignment until death of any cause or date of last follow-up. Likewise, PFS was defined as time until progression or death or date of last follow-up. OS and PFS were estimated by the Kaplan–Meier method and statistically evaluated by a log rank test. Multivariable Cox regression models were calculated to analyze the impact of residual tumor on survival, adjusting for other predictors of survival.24,25 To control, for example, for possible time trends confounding the relation between residual tumor and survival, we adjusted in the multivariable analysis for the patient’s allocation to one of the studies (OVAR-3 to -7). Hazard ratios are presented with their 95% confidence intervals (95% CI). P values given are to be interpreted as descriptive accounting for the exploratory nature of this study. Calculations were performed by SPSS software, version 16.0 (SPSS, Chicago, IL).

Results

The three randomized phase III trials included 3388 patients with advanced epithelial ovarian cancer between 1995 and 2002. A total 573 patients had FIGO stage IV disease. Of these, 214 patients (37.3%) had malignant pleural effusions, 146 patients (25.5%) had parenchymal hepatic metastases, and in 213 patients (37.2%), other sites were involved in distant metastatic spread (e.g., extra-abdominal lymph nodes, penetrating abdominal wall metastases, pulmonary metastases, or bone lesions). Sixty-three patients (10.9%) had multiple and 510 patients (89.1%) had single sites of distant metastases defining FIGO stage IV disease.

Patient characteristics are displayed for the subgroup with FIGO stage IV disease in Table 1. The median age of patients was 59.0 years (range, 19–83 years). A total of 68.2% showed a histologically serous cell type, whereas 6.9% had a mucinous subtype. Residual disease (intra- and extraperitoneal) after surgery was microscopically seen in only 12.3%. In 29.3% a residual tumor of 1 to 10 mm and in 58.4% a residual tumor of >1 cm were reported.

Table 1 Univariate analysis of prognostic factors in FIGO stage IV disease (n = 573)
Full size table

The different platinum-taxane regimens did not show any marked impact on OS, as already stated in the original analyses of all three studies (data not shown).

Within the OVAR-3 trial, 69 patients with FIGO stage IV disease received carboplatin-paclitaxel and 64 patients cisplatin-paclitaxel; 7 versus 6 patients underwent complete resection, respectively. In the OVAR-5 trial, 112 patients with FIGO stage IV disease were randomized in the carboplatin-paclitaxel arm and 106 in the carboplatin-paclitaxel-epirubicin arm. Fourteen versus 12 patients underwent complete resection, whereas 61 versus 63 patients had a residual tumor of >1 cm after surgery. Within the OVAR-7 trial, 104 patients with FIGO stage IV disease were enrolled onto the carboplatin-paclitaxel arm and 118 patients in the carboplatin-paclitaxel-topotecan arm. In each arm, 15 patients had complete resection.

In each study, the difference in the proportion of patients with zero residual tumor between study arms is not statistically significant (OVAR-3, P = 0.88; OVAR-5, P = 0.79; OVAR-7, P = 0.71). Also, a trend in this proportion with later recruitment period (OVAR-3, from 1995 to 1997, 9.8%; OVAR-5, from 1997 to 1999, 11.9%; OVAR-7, from 1999 to 2002, 13.5%) is not significant (P = 0.30). For all three trials (OVAR-3, -5, and -7), we found no relation between residual tumor and the number of applied chemotherapy cycles. The OVAR-3/-5/-7 study groups are therefore considered sufficiently similar to be combined for this analysis.

OS was statistically significantly reduced in FIGO stage IV disease (median OS, 26.1 months) in comparison to FIGO stage IIIC disease (median OS, 40.5 months). This was observed for all three localization subgroups of FIGO stage IV disease (Fig. 1). Analogous relations were found for PFS periods (median PFS, 12.6 vs. 17.0 months).

Fig. 1
figure 1

Overall survival of International Federation of Gynecology and Obstetrics (FIGO) stage IIIC disease in comparison to different sites of distant metastatic spread (malignant pleural effusion, parenchymal liver metastases, and any other site of distant metastatic spread)

Full size image

Multiple sites of distant metastases showed a reduced median OS of 20.2 months (95% CI, 14.0–26.4) in comparison to single-site metastases (median OS: 26.8 months; 95% CI, 23.7–29.9; P = 0.009). However, median PFS was not significantly different (multiple sites, 11.5 months, vs. single site, 12.8 months; P = 0.103).

Complete resection with no macroscopic evidence of residual tumor was achieved in 70 patients with FIGO stage IV disease (12.3% of the whole study population with FIGO stage IV disease). Complete resection was associated with significantly improved OS rates in comparison to the cohort with visible residual tumor, including minimal residual disease (P < 0.0001) (Fig. 2). Median OS period after complete debulking in FIGO stage IV disease was 54.6 months (95% CI, 35.7–73.6), whereas the subgroup of patients with residual tumors of 1 to 10 mm showed a median OS of 25.8 months (95% CI, 21.1–30.4). Patients with residual tumor of >1 cm had a median survival time of 23.9 months (95% CI, 20.7–27.2). Accordingly, the median PFS period after complete debulking was 19.1 months (95% CI, 15.6–22.4 months), whereas the subgroup of patients with residual tumor of 1 to 10 mm showed a median PFS of 13.6 months (95% CI, 12.0–15.4). Patients with residual tumor of >1 cm had a median PFS of 11.3 months (95% CI, 10.2–12.3).

Fig. 2
figure 2

Overall survival for cases with no residual tumor vs. residual tumor of 1–10 mm and >1 cm in International Federation of Gynecology and Obstetrics (FIGO) stage IIIC and FIGO stage IV disease

Full size image

A residual tumor size of >5 cm showed no further diminished PFS and OS when compared with all other subgroups with any residual tumor (data not shown). In addition, the subgroup of patients with macroscopically complete resection had an improved PFS compared to all other subgroups with residual tumor.

Univariate analysis revealed complete resection, Eastern Cooperative Oncology Group (ECOG) performance status (ECOG 0 vs. >0), peritoneal carcinomatosis, and mucinous histological subtype to be statistically significant prognostic factors for OS and for PFS (Table 1). The mucinous histological subtype predicted a poorer OS in comparison to other histological subtypes (P < 0.0001). In the univariate analysis, multiple sites of distant metastases and parenchymal liver metastases were the only sites to show a statistically significant negative impact on the period of survival. Multivariable analyses showed residual tumor (microscopic vs. visible residual tumor), number of metastatic sites (multiple vs. single site), ECOG performance status, and mucinous histological subtype to be independent prognostic factors for OS. Specific disease localization defining FIGO stage IV disease was not found to be a statistically significant prognosticator. Residual tumor (microscopic vs. visible residual tumor), peritoneal carcinomatosis, and mucinous histological subtype were independent prognosticators of PFS (Table 2). In the multivariable models embodying prognostic factors and patient allocation to study, there was neither a significant difference in PFS between studies (OVAR-3, HR 1.0; OVAR-5, HR 1.08, 95% CI, 0.84–1.39, P = 0.549; OVAR-7, HR 1.05, 95% CI, 0.82–1.35, P = 0.686) nor in OS (OVAR-3, HR 1.0; OVAR-5, HR 1.14, 95% CI, 0.87–1.48, P = 0.35; OVAR-7, HR 1.10, 95% CI, 0.84–1.44, P = 0.482).

Table 2 Multivariable analysis of prognostic factors in FIGO stage IV disease
Full size table

Statistically significantly improved OS periods were registered after complete resection in comparison to residual tumor for the subgroup with intrahepatic metastases (P = 0.013). Patients with malignant pleural effusion also showed an improved survival after complete resection of all metastases; however, this remained statistically insignificant (P = 0.083).

Discussion

This exploratory analysis of FIGO stage IV ovarian cancer was based on a large series of patients prospectively enrolled onto three consecutive, randomized phase III trials. All patients were selected by almost identical inclusion criteria and were treated at member centers of two European cooperative study groups. We believe that series such as these and similar reports on exploratory analyses of large study databases may help evaluate the impact of prognostic factors in patients with advanced ovarian cancer.10,15,26 Another strength of our series is that all patients received platinum- and paclitaxel-based chemotherapy within 6 weeks after cytoreduction surgery, thus eliminating any variables associated with heterogenous chemotherapy.

Surgical outcome after maximum cytoreductive surgery is one of the most powerful determinants of survival in patients with advanced ovarian cancer.22 This has been well accepted for stage III disease; however, little information is available for the subgroup of patients with FIGO stage IV disease. Therefore, this study, which represents what is to our knowledge the largest investigation of stage IV epithelial ovarian cancer patients, may add valuable information to the already published smaller series that reported on prognostic factors in stage IV disease.5–8,15,17,18

Our results may not be valid for patients receiving neoadjuvant chemotherapy followed by interval debulking surgery. Some retrospective series have already suggested that the role of complete macroscopic resection after neoadjuvant chemotherapy might not have a comparable impact as after primary debulking.27 Despite an increased rate of optimal cytoreduction, no improved overall and PFS outcomes compared with standard primary debulking were present after neoadjuvant chemotherapy.28,29

Complete tumor resection to microscopic residuals greatly improved overall and PFS in our cohort. In contrast to the retrospective analysis of Winter et al., we found no difference in survival for patients with residual tumor of >5 cm compared to patients with residual tumor of <5 cm.15 The relative clinical impact of complete debulking observed in patients with FIGO stage IV disease was comparable to the impact already reported for stage III disease.22 In FIGO stage IIIc disease, OS was 78.5 months in cases with macroscopically complete resection and 35.8 months in cases with a residual tumor of 1 to 10 mm. In FIGO stage IV disease, OS was 54.6 months in cases with macroscopically complete resection and 25.8 months in cases with a residual tumor of 1 to 10 mm. This implies that patients with FIGO stage IV disease and complete resection have a better outcome than patients with residual tumor with FIGO stage IIIc disease. By analogy with OS, PFS was 35.0 months in cases with macroscopically complete resection and 15.9 months in patients with a residual tumor of 1 to 10 mm in FIGO stage IIIc. In FIGO stage IV disease, PFS was 19.1 months in patients with macroscopically complete resection and 13.6 months in patients with a residual tumor of 1 to 10 mm.

Therefore, clinicians have more justification for aggressive surgery in patients with stage IV disease. This observation did not surprise us because we thought that FIGO classification is a system of practical value that is based on empirical data, rather than an attempt at describing tumor biology. FIGO arbitrarily defines some localizations as compatible with stage III, whereas others define stage IV (e.g., inguinal nodes define FIGO stage III while affected lymph nodes in the mesenterium of the digestive tract define stage FIGO stage IV). Although some metastases defining FIGO stage IV (e.g., brain metastases) may reflect a different biologic behavior, others (e.g., Sister Joseph node) may only reflect a continuum and more advanced growth, the latter possibly depending on the time point of diagnosis rather than on biologic behavior.

In contrast to other investigators, we found in our exploratory analysis of FIGO stage IV disease no marked difference in survival between groups with residual tumor of 1 to 10 mm and of >1 cm; this also applied to cutoffs of 2 or 5 cm.5,7 An explanation could be a different biology in patients with FIGO stage IV disease independent of localization of distant metastases. It is important to note that in ovarian cancer, hematogenous metastases are rare in comparison to other solid tumors.

Apart from the chosen surgical procedure, the biological nature of the disease is a further important factor in complete tumor resection.

Within our study population, only 12.3% of the patients with FIGO stage IV disease had macroscopically complete resection. We found no correlation with the surgery-performing institutions; however, only small patient numbers were available. All patients were treated in member centers of cooperative study groups and received systemic treatment comprising the currently regarded most effective drugs in ovarian cancer, namely platinum and paclitaxel. However, surgical outcome reflects a broad variety of hospitals participating in these studies.

The importance of biologic prognosticators was underlined by our observation that histological cell type had an tremendous prognostic impact irrespective of the FIGO stage. As already indicated in our main analysis, in the entire population of patients with advanced ovarian cancer including stages FIGO IIb–IIIc, mucinous histology was a very strong prognostic factor not correctable by any therapeutic intervention.22 The particularly poor outcome in the mucinous adenocarcinoma subgroup has been reported by others as well and may indicate a need for different studies addressing specific questions for this subtype.30 The data now available suggest that future trials should at least stratify for histology to avoid bias by maldistribution of mucinous tumors.

The weakness of our study lies in the retrospective design, which cannot exclude all possible selection bias. However, the three basic studies were stratified for FIGO stage, and stage IV defined one strata, thus assuring balanced randomization into the studies. Furthermore, even this large series provided only small subgroups of patients with specific patterns of distant metastases. Therefore, we were able to analyze only two distinct subgroups of patients with FIGO stage IV disease, namely those with parenchymal liver metastases or with malignant pleural effusion. Future analyses on larger trial data bases are necessary to evaluate other specific sites of disease.