SEARCH

SEARCH BY CITATION

Keywords:

  • epithelial ovarian cancer;
  • CA 125 tumor marker;
  • Gynecologic Oncology Group;
  • progression-free survival

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Conflict of Interest Disclosures
  7. References

BACKGROUND:

The objective of the current study was to determine the prognostic significance of a pretreatment serum CA 125 level in patients with advanced epithelial ovarian carcinoma (EOC) who received treatment with a standard chemotherapy regimen.

METHODS:

Patients with International Federation of Gynecology and Obstetrics stage III/IV ovarian carcinoma who were on 1 of 7 Gynecologic Oncology Group (GOG) phase 3 trials and received treatment with a standard regimen of intravenous cisplatin and paclitaxel were included. A Cox regression model was used to assess the impact of CA 125 levels drawn before the initiation of chemotherapy on progression-free survival (PFS) both overall and by subgroup, including surgical debulking status, disease stage, and histologic subtype.

RESULTS:

In total, 1299 patients who were on the cisplatin/paclitaxel arms of the GOG trials were eligible. The median CA 125 level was 246 U/mL. Only 7.6% of patients had a normal CA 125 level (≤35 U/mL). The lowest median CA 125 level was observed in the group with mucinous tumors; however, 69% of women who had mucinous tumors had abnormal CA 125 levels. Shorter PFS was observed with increasing CA 125 and persisted in multivariate analysis. Overall and in the serous subgroup, a 1-fold increase in CA 125 level was associated with a 7% increase in the hazard of disease progression (P < .001). This association was even more pronounced in patients who had stage III disease that was debulked to microscopic disease (15%; P = .003) and in patients who had endometrioid tumors (17%; P = .001).

CONCLUSIONS:

A normal CA 125 level in the setting of advanced EOC was rare even after surgical debulking. The pretreatment CA 125 level was an independent predictor of PFS in patients with advanced EOC who received a standard chemotherapy regimen, particularly in the setting of disease that was debulked to a microscopic residual and in the serous or endometrioid subtypes. Cancer 2009. © 2009 American Cancer Society.

CA 125 is the serum tumor marker associated most closely with epithelial ovarian carcinoma (EOC). Originally described by Bast et al in 1981, it is an antigenic determinant on a high-molecular-weight glycoprotein that recently was described as mucin 16 and can be recognized by the murine monoclonal antibody OC-125.1, 2 In subsequent studies, CA 125 values >35 U/mL were documented in 80% to 85% of women who were diagnosed with ovarian cancer.3, 4 For women with advanced disease, the rate of elevated CA 125 is even higher at >90%.4

The prognostic significance of CA 125 is well recognized in EOC, but to our knowledge, the majority of the patients studied to date have had tumors with serous histology. The prognostic value of CA 125 for the other histologic subtypes of EOC has not been well documented, especially when stage and surgical debulking status are taken into consideration. In addition, to our knowledge, the prognostic significance of a normal CA 125 level in the setting of advanced EOC has not been evaluated previously. Some of the difficulty arises from the lack of understanding of why some tumors are associated with an elevated CA 125 level whereas others are not. The most commonly cited reason for a normal CA 125 level is the mucinous histologic subtype; however, 69% of these tumors have been associated with abnormal CA 125 levels.4

The objective of this study was to evaluate whether the pretreatment CA 125 level can be used to predict disease progression in patients with advanced EOC, particularly when stage, debulking status, and histologic subtype are taken into consideration. Conducted as an ancillary data project of the Gynecologic Oncology Group (GOG), the study incorporated patients from 7 closed GOG phase 3 protocols that used a standard regimen of intravenous cisplatin and paclitaxel in 1 treatment arm.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Conflict of Interest Disclosures
  7. References

A retrospective review of data from 7 prospective, randomized clinical trials conducted by the GOG, including protocols 111, 114, 132, 152, 158, 162, and 172, was performed. The patients who were included in the current analysis had undergone primary surgical cytoreduction followed by 6 cycles of intravenous paclitaxel at a dose of 135 mg/m2 over 24 hours and cisplatin at a dose of 75 mg/m2. Although patients on the experimental arm of GOG 152 received the same chemotherapy regimen divided before and after secondary cytoreductive surgery, they were excluded in an attempt to maintain the most consistent treatment possible. Likewise, the patients on GOG 162 who received the 96-hour paclitaxel regimen also were excluded. For information regarding patient eligibility, treatment, and clinical outcomes, please refer to previous publications.5-11

CA 125 collection after surgery but before chemotherapy initiation was requested in all of the protocols. Demographic information, disease stage, debulking status (optimal with microscopic residual disease, optimal with macroscopic residual disease, or suboptimal), tumor grade, and histologic subtype were evaluated. The associations between CA 125 and these clinical characteristics were assessed by an analysis of variance in which CA 125 values were transformed to a logarithm scale.

The endpoint of the study was progression-free survival (PFS), which we defined as the time interval from study entry to disease recurrence or death. PFS by CA 125 stratum was estimated by using the Kaplan-Meier method. CA 125 levels were divided into 6 categories: ≤35 U/mL, 36 to 100 U/mL, 101 to 200 U/mL, 201 to 500 U/mL, 501 to 1000 U/mL, and >1000 U/mL. This classification took into account the most common definition of a normal CA 125 level (≤35 U/mL) and allowed comparable numbers of patients in each category. The relative risk of disease progression was estimated using a Cox model that was adjusted for age, GOG performance status, stage/debulking status, histology, and tumor grade. CA 125 was analyzed either as a categoric variable or as a continuous variable. When it was analyzed as a categoric variable, the relative risk was reported compared with the normal group (CA 125 ≤35 U/mL); when it was analyzed as a continuous variable, the relative risk was reported for a 1-unit increase in log2(CA 125), which can be interpreted as the relative risk with a 1-fold increase in the raw CA 125 level. The linearship between log2(CA 125) and the risk of disease progression was assessed using the graphic method suggested by Hosmer and Lemeshow.12 Briefly, the patients were divided into 10 groups based on their CA 125 level, and each group had an equal number of patients. Design variables were created for each group and were included in the Cox model. The regression coefficients against the group midpoints were graphed using a smoothing method. Subgroup analysis by stage/debulking status (stage III, optimal debulking with microscopic residual disease; stage III, optimal debulking with macroscopic residual disease; stage III, suboptimal debulking; and stage IV) and by histologic subtype (serous, endometrioid, mucinous, clear cell, and other) was performed using the same method with CA 125 analyzed as continuous variable. All statistical analyses were performed using SAS statistical software (version 9.1; SAS, Inc, Cary, NC) and P values are reported as 2-sided.

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Conflict of Interest Disclosures
  7. References

From a total of 1570 patients who were treated on the cisplatin/paclitaxel arms of the 7 GOG protocols, 1299 patients (83%) had pretreatment CA 125 values available and were included in this analysis (Table 1). The exact timing of the postoperative CA 125 value (ie, which postoperative day) was not specified. Nine values were obtained preoperatively with no repeat value reported after surgery but before chemotherapy. These values all were <35 U/mL and were included in the analysis, because they represented pretreatment values that already were normal before surgery.

Table 1. Gynecologic Oncology Group Protocols and Study Populations
GOG ProtocolPatient PopulationTotal No. of Patients on the Cisplatin 75 mg/m2 Plus Paclitaxel 135 mg/m2 Over 24 Hours Treatment ArmNo. of Patients With Pretreatment CA 125 (%)
  1. GOG indicates Gynecologic Oncology Group.

  2. Suboptimal disease was defined as residual disease that measured >1 cm in greatest dimension for all studies except GOG 162, in which suboptimal disease was defined as retroperitoneal disease that measured >1 cm or intraperitoneal disease that measured ≤1 cm. Optimal disease defined as residual disease that measured ≤1 cm for all studies. Patients who did not have a pretreatment CA 125 level available, who underwent interval debulking (GOG 152), who were treated with a 96-hour infusion (GOG 162), or who received intraperitoneal treatment (GOG 172) were not included in the current analysis. See McGuire 1996,5 Markman 2001,6 Muggia 2000,7 Rose 2004,8 Ozols 2003,9 Spriggs 2004,10 and Armstrong 2006.11

111Suboptimal, stage III/IV184163 (89)
114Optimal, stage III226179 (79)
132Suboptimal, stage III/IV201181 (90)
152Suboptimal, stage III /IV209169 (81)
158Optimal, stage III400343 (86)
162Suboptimal, stage III /IV14094 (67)
172Optimal, stage III210170 (81)
Total 15701299 (83)

The median pretreatment CA 125 level was 246 U/mL, whereas only 99 patients (7.6%) had a normal level (≤35 U/mL). The characteristics of these 99 patients are summarized in Table 2. Seventy-nine patients had residual disease that measured ≤1 cm after debulking surgery, whereas 20 patients had either suboptimal debulking or stage IV disease. The majority of patients with normal pretreatment CA 125 levels had serous histology (64.6%); however, among the histologic subtypes, patients who had tumors with mucinous histology had the largest percentage of normal pretreatment CA 125 levels (30.8%). Patients who had tumors with mucinous histology had a median pretreatment CA 125 level of 99 U/mL, which was the lowest among the histologic subtypes but still in the abnormal range (Table 3).

Table 2. Characteristics of Patients With Normal Pretreatment CA 125 Levels
CharacteristicDistribution of 99 Patients With Normal Pretreatment CA 125, %Percentage of Patients in Each Subgroup With Normal Pretreatment CA 125
Stage  
 III: Optimal microscopic40.416
 III: Optimal macroscopic39.38.8
 III: Suboptimal15.23.6
 IV5.12.7
Histologic subtype  
 Serous64.66.8
 Endometrioid6.15.2
 Mucinous8.130.8
 Clear cell38.1
 Other18.29.8
Table 3. Patient Characteristics and Pretreatment CA 125 Levels
CharacteristicNo. of PatientsMedian CA 125 (25th to 75th Percentile), U/mLLog2 (CA 125), Mean±SDP*
  • SD indicates standard deviation; log2(CA 125), the relative risk with a 1-fold increase in the raw CA 125 level; GOG, Gynecologic Oncology Group.

  • *

    Reported from an analysis of variance based on log2(CA 125).

Age, y   .02
 ≤50365202 (85-598)7.85±2.11 
 51-60370278 (118-903)8.35±2.25 
 61-70382256 (113-722)8.19±2.13 
 >70182224 (105-667)8.07±2.12 
Race   .17
 White1157251 (103-740)8.15±2.14 
 Black74142 (60-576)7.66±2.56 
 Other68239 (151-761)8.12±2.09 
GOG performance status   <.001
 0509203 (84-601)7.86±2.16 
 1650258 (112-840)8.23±2.14 
 2140356 (150-840)8.58±2.22 
Stage   <.001
 III: Optimal microscopic250117 (54-221)6.86±1.78 
 III: Optimal macroscopic442203 (86-485)7.77±1.91 
 III: Suboptimal420407 (170-1393)8.85±2.21 
 IV187443 (170-1427)9.01±2.10 
Histology   .007
 Serous936256 (103-759)8.17±2.15 
 Endometrioid117239 (114-561)8.12±2.08 
 Mucinous2699 (26-336)6.81±2.38 
 Clear cell37154 (87-390)7.43±1.74 
 Other183247 (102-840)8.22±2.23 
Tumor grade   <.001
 1115120 (55-430)7.19±2.04 
 2489227 (96-723)8.08±2.21 
 3658296 (117-838)8.35±2.12 
 Clear cell37154 (87-390)7.43±1.74 
Total1299246 (101-741)8.12±2.17 

CA 125 levels differed significantly by age group, performance status, stage/debulking status, histology and tumor grade (Table 3). A lower pretreatment CA 125 level was associated with age ≤50 years and normal performance status (GOG performance status 0).

With a median follow-up of 39 months for all patients, 1132 patients (87.1%) had disease progression (either disease recurrence or death). With CA 125 levels stratified into 6 categories, Kaplan-Meier survival estimates revealed an association of shorter PFS with increasing pretreatment CA 125 except for comparable PFS in the 200 to 500 U/mL and 500 to 1000 U/mL strata (Fig. 1). By using the ≤35 U/mL stratum as the referent, the association of pretreatment CA 125 level with disease progression was evaluated in a Cox model. Hazards ratios (HRs) were adjusted for age, performance status, stage/debulking status, histology, and tumor grade (Fig. 2A). As pretreatment CA 125 level increased, the risk of disease progression also increased until, at levels >1000 U/mL, there was a 71% increase in the risk of disease progression compared with a normal level. When CA 125 was analyzed as a continuous variable, the logarithm scale for the HR was approximately linear to the logarithm scale for the pretreatment CA 125 level (Fig. 2B), with an 11% increase for each doubling of the pretreatment CA 125 level (HR, 1.11; 95% confidence interval [95% CI], 1.08-1.14 [P < .001]). When the model was adjusted for covariates, the increased risk of disease progression was 7% for each doubling of the pretreatment CA 125 level (HR, 1.07; 95% CI, 1.04-1.10 [P < .001]). A similar analysis of the relation between pretreatment CA 125 and overall survival also indicated a significant association (data not shown).

thumbnail image

Figure 1. Estimate for progression-free survival (PFS) by CA 125 level.

Download figure to PowerPoint

thumbnail image

Figure 2. Association between pretreatment CA 125 level and disease progression. (A) Adjusted hazards ratios (HRs) from multivariate analysis. HRs were estimated by using a Cox model that was adjusted for age, Gynecologic Oncology Group (GOG) performance status, stage/debulking status, histology, and tumor grade; CA 125 was defined as either a categoric variable or a continuous variable (based on the logarithm). For categoric variables, the HR was interpreted compared with the normal group (CA-125 level ≤35 U/mL); for continuous variables, the HR was interpreted as a 1-fold increase in the raw CA 125 level or log2(CA 125). (B) Assessment of linearship. 95% CI indicates 95% confidence interval.

Download figure to PowerPoint

Cox modeling identified an interactive effect between the pretreatment CA 125 level and stage/debulking status on the risk of disease progression (test for interaction; P = .03). Further analyses were conducted to assess the association between pretreatment CA 125 and PFS for each subgroup. Figure 3A shows the adjusted risk of disease progression with a 1-fold increase in pretreatment CA 125 according to stage/debulking status. The extent of association decreased, moving from patients with stage III, low-residual disease, to patients with stage IV disease. For example, a 1-fold increase in the pretreatment CA 125 level was associated with a 15% increase in the risk of disease progression for patients with stage III disease debulked to a microscopic residual (HR, 1.15; 95% CI, 1.05-1.26 [P = .003]), whereas the increased risk was only 6% for patients who had stage IV disease (HR, 1.06; 95% CI, 0.99-1.14 [P = .10]).

thumbnail image

Figure 3. (A) Association between pretreatment CA 125 level and disease progression by stage/surgical debulking status. The adjusted hazards ratios are reported as a 1-fold increase in the pretreatment CA 125 level. Stage/debulking status was reported for stage III disease that was debulked optimally to a microscopic residual, for stage III disease that was debulked optimally to a residual ≤1 cm, for stage III disease that was debulked suboptimally, and for stage IV disease. (B) Association between pretreatment CA 125 level and disease progression by histologic subtype. Adjusted hazards ratios are reported as a 1-fold increase in the pretreatment CA 125 level.

Download figure to PowerPoint

No statistically significant interaction between pretreatment CA 125 and histologic subtype was identified overall (P = .241). However, the adjusted risk of disease progression with a 1-fold increase in pretreatment CA 125 depicted in Figure 3B indicated that the association was more evident for the endometrioid subtype (HR, 1.17; 95%CI, 1.06-1.28 [P = .001]) than for the serous subtype (HR, 1.07; 95% CI, 1.04-1.11 [P < .001]). The correlation between the pretreatment CA 125 level and the risk of disease progression for mucinous and clear cell subtypes was not statistically significant, most likely because of the small sample sizes for those 2 groups.

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Conflict of Interest Disclosures
  7. References

This large cohort of women with advanced EOC was treated on randomized cooperative group protocols with a consistent chemotherapy regimen and underwent centralized data review. The primary limitation of the current study was the retrospective nature of the data collection. The timing of CA 125 collection was not specified by the protocols, leading to a mix of specimens collected preoperatively, immediately postoperatively, or later in the postoperative course but before chemotherapy was initiated. CA 125 levels were recorded as raw values without the local laboratory reference range. The upper limit of normal on current assays ranges from 21 U/mL to 40 U/mL. For the purposes of the current analysis, the commonly accepted definition of normal CA 125 (≤35 U/mL) was used. This analysis also was limited by the use of a single pretreatment value rather than repeated pretreatment values. Because of the lack of paired preoperative and postoperative CA 125 values, the immediate impact of surgery and debulking on pretreatment CA 125 values could not be assessed. Previous research suggests that surgical trauma may temporarily elevate the CA 125 level in patients who have benign disease and a previously normal level, whereas debulking ovarian cancer generally leads to a decrease in a previously elevated CA 125 level.13 An additional limitation is the small number of mucinous and clear cell tumors that were included. Despite its multi-institutional nature, this analysis only included 26 patients with mucinous EOC and 37 patients with clear cell EOC, emphasizing the difficulty of studying the rare histologic subtypes even in a cooperative group setting.

Although it is believed generally that CA 125 levels remain normal in 15% to 20% of women with EOC, that rate drops to <10% in women with advanced-stage disease.4 A recent analysis of data from a Southwest Oncology Group trial indicated that 7 of 101 patients (6.9%) who had suboptimally debulked, advanced EOC had a pretreatment CA 125 level ≤35 U/mL.14 Similarly, in the current analysis, which included a mix of optimally and suboptimally debulked patients, only 7.6% of patients had a pretreatment CA 125 level ≤35 U/mL, confirming the original studies from the 1980s.

In addition, these data suggest that histology and resectability alone do not explain the absence of CA 125 elevation in the setting of advanced EOC. Specifically, although the mucinous group had the highest percentage of patients with a CA 125 level ≤35 U/mL among the histologic subtypes (30.8%), patients with mucinous tumors accounted for only 8.1% of all patients with normal CA 125 levels. Furthermore, 20.3% of patients with normal CA 125 levels had suboptimally debulked stage III or IV disease.

A recent GOG analysis that included patients from 6 of the studies that also were included in this study indicated that age, performance status, tumor residual after debulking surgery, and tumor histology were independent prognostic factors for stage III EOC.15 Our regression analysis included clinical factors, such as patient age, GOG performance status, stage/debulking status, histology, and tumor grade, to control for their impact on the association of CA 125 level with outcome. In general, these clinical factors may contribute more to patient outcomes, making the CA 125 contribute comparatively less. However, in the subgroup analyses in which the most important clinical factors, such as tumor residual or histology, was controlled, the relative prognostic value of CA 125 level was more apparent among patients who had favorable clinical factors. This may help explain the higher significance of the pretreatment CA 125 association with PFS in the microscopic residual, serous histology, and endometrioid histology groups, because these groups usually have a better prognosis than women with stage IV disease, suboptimal debulking, and unfavorable clear cell or mucinous histology. Extending the evaluation of pretreatment CA 125 to women with early-stage ovarian cancer would add to the recent analysis of prognostic factors in this group; as a population with an even more favorable prognosis than stage III patients with microscopic residual and serous or endometrioid histology, CA 125 may be a significant, independent predictor of outcome.16

Although many clinical studies have concentrated on the use of serial measurements of CA 125 during chemotherapy for prognosis, the pretreatment CA 125 level remains of interest, because it has the potential to predict prognosis even before the start of chemotherapy when it is used together with other clinical factors. In addition to stage, tumor residual, histologic subtype, tumor grade, age, and GOG performance status, the current analysis suggests that pretreatment CA 125 can be considered a prognostic factor in patients with EOC. The use of a large population that was treated similarly makes the estimate of relative risk more accurate. After adjusting for covariates, we observed that a 1-fold increase in the pretreatment CA 125 level was associated with a 7% increase in the risk of progression in patients with advanced EOC. Knowledge of poor prognostic factors at the start of chemotherapy may help clinicians and patients to prioritize their participation in protocol therapy that investigates novel agents in the treatment of EOC.

Conflict of Interest Disclosures

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Conflict of Interest Disclosures
  7. References

Supported by National Cancer Institute grants to the Gynecologic Oncology Group Administrative Office (CA-27469) and the Gynecologic Oncology Group Statistical Office (CA-37517).

References

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Conflict of Interest Disclosures
  7. References
  • 1
    Bast RCJr,Feeney M,Lazarus H, et al. Reactivity of a monoclonal antibody with human ovarian carcinoma. J Clin Invest. 1981; 68: 1331-1337.
  • 2
    Yin BW,Lloyd KO. Molecular cloning of the CA125 ovarian cancer antigen: identification as a new mucin, MUC16. J Biol Chem. 2001; 20: 27371-27375.
  • 3
    Bast RCJr,Klug TL,St John E, et al. A radioimmunoassay using a monoclonal antibody to monitor the course of epithelial ovarian cancer. N Engl J Med. 1983; 309: 883-887.
  • 4
    Jacobs I,Bast RJJr. The CA 125 tumour-associated antigen: a review of the literature. Hum Reprod. 1989; 4: 1-12.
  • 5
    McGuire WP,Hoskins WJ,Brady MF, et al. Cyclophosphamide and cisplatin compared with paclitaxel and cisplatin in patients with stage III and stage IV ovarian cancer. N Engl J Med. 1996; 334: 1-6.
  • 6
    Markman M,Bundy B,Alberts D, et al. Phase III trial of standard-dose intravenous cisplatin plus paclitaxel versus moderately high-dose carboplatin followed by intravenous paclitaxel and intraperitoneal cisplatin in small-volume stage III ovarian carcinoma: an intergroup study of the Gynecologic Oncology Group, Southwestern Oncology Group, and Eastern Cooperative Oncology Group. J Clin Oncol. 2001; 19: 1001-1007.
  • 7
    Muggia FM,Braly PS,Brady MF, et al. Phase III randomized study of cisplatin versus paclitaxel versus cisplatin and paclitaxel in patients with suboptimal stage III or IV ovarian cancer: a Gynecologic Oncology Group study. J Clin Oncol. 2000; 18: 106-115.
  • 8
    Rose PG,Nerenstone S,Brady MF, et al. Secondary surgical cytoreduction for advanced ovarian carcinoma. N Engl J Med. 2004; 351: 2489-2497.
  • 9
    Ozols RF,Bundy BN,Greer BE, et al. Phase III trial of carboplatin and paclitaxel compared with cisplatin and paclitaxel in patients with optimally resected stage III ovarian cancer: a Gynecologic Oncology Group study. J Clin Oncol. 2003; 21: 3194-3200.
  • 10
    Spriggs D,Brady MF,Rubin SC, et al. A phase III randomized trial of cisplatin and paclitaxel administered by either 24 hour or 96 hour infusion in patients with selected stage III or stage IV epithelial ovarian cancer [abstract]. Proc Am Soc Clin Oncol. 2004; 23: 449. Abstract 5004.
  • 11
    Armstrong DK,Bundy B,Wenzel L, et al. Intraperitoneal cisplatin and paclitaxel in ovarian cancer. N Engl J Med. 2006; 354: 34-43.
  • 12
    Hosmer D,Lemeshow S. Applied Survival Analysis: Regression Modeling of Time to Event Data. New York: John Wiley & Sons; 1999.
  • 13
    Talbot RW,Jacobsen DJ,Nagorney DM, et al. Temporary elevation after abdominal surgical treatment for benign disease and cancer. Surg Gynecol Obstet. 1989; 168: 407-412.
  • 14
    Markman M,Federico M,Liu PY, et al. Significance of early changes in the serum CA-125 antigen level on overall survival in advanced ovarian cancer. Gynecol Oncol. 2006; 103: 195-198.
  • 15
    Winter ME3rd,Maxwell GL,Tian C, et al. Gynecologic Oncology Study Group. Prognostic factors for stage III epithelial ovarian cancer: a Gynecologic Oncology Group study. J Clin Oncol. 2007; 24: 3621-3627.
  • 16
    Chan JK,Tian C,Monk BJ, et al. Gynecologic Oncology Group. Prognostic factors for high-risk early-stage epithelial ovarian cancer: a Gynecologic Oncology Group study. Cancer. 2008; 112: 2202-2210.