The following Gynecologic Oncology Group member institutions participated in this study: University of Alabama at Birmingham, Duke University Medical Center, Abington Memorial Hospital, University of Rochester Medical Center, Walter Reed Army Medical Center, Wayne State University, University of Minnesota Medical School, Emory University Clinic, University of Mississippi Medical Center, Colorado Gynecologic Oncology Group P.C., University of California at Los Angeles, University of Washington, University of Pennsylvania Cancer Center, Milton S. Hershey Medical Center, Georgetown University Hospital, University of Cincinnati, University of North Carolina School of Medicine, University of Texas Southwestern Medical Center at Dallas, Indiana University School of Medicine, Wake Forest University School of Medicine, Albany Medical College, University of California Medical Center at Irvine, Tufts-New England Medical Center, Rush-Presbyterian-St. Luke's Medical Center, SUNY Downstate Medical Center, University of Kentucky, Community Clinical Oncology Program, The Cleveland Clinic Foundation, Johns Hopkins Oncology Center, State University of New York at Stony Brook, Eastern Pennsylvania GYN/ONC Center, P.C., Washington University School of Medicine, Cooper Hospital/University Medical Center, Columbus Cancer Council, University of Massachusetts Medical School, Fox Chase Cancer Center, Medical University of South Carolina, Women's Cancer Center, University of Oklahoma, University of Virginia Health Sciences Center, University of Chicago, Tacoma General Hospital, Thomas Jefferson University Hospital, Case Western Reserve University, Tampa Bay Cancer Consortium, North Shore University Hospital, Gynecologic Oncology Network, Oregon Health Sciences University, University of Southern California at Los Angeles, University of Miami School of Medicine, Stanford University Medical Center, Eastern Virginia Medical School, University of Arizona Health Science Center, Mayo Clinic, and Long Island Jewish Medical Center.
A study was undertaken to use the 2-tier system to reclassify the grade of serous ovarian tumors previously classified using the International Federation of Gynecology and Obstetrics (FIGO) 3-tier system and determine the progression-free survival (PFS) and overall survival (OS) of patients treated on Gynecologic Oncology Group (GOG) Protocol 158.
The authors retrospectively reviewed demographic, pathologic, and survival data of 290 patients with stage III serous ovarian carcinoma treated with surgery and chemotherapy on GOG Protocol 158, a cooperative multicenter group trial. A blinded pathology review was performed by a panel of 6 gynecologic pathologists to verify histology and regrade tumors using the 2-tier system. The association of tumor grade with PFS and OS was assessed.
Of 241 cases, both systems demonstrated substantial agreement when combining FIGO grades 2 and 3 (overall agreement, 95%; kappa statistic, 0.68). By using the 2-tier system, patients with low-grade versus high-grade tumors had significantly longer PFS (45.0 vs 19.8 months, respectively; P = .01). By using FIGO criteria, median PFS for patients with grade 1, 2, and 3 tumors was 37.5, 19.8, and 20.1 months, respectively (P = .07). There was no difference in clinical outcome in patients with grade 2 or 3 tumors in multivariate analysis. Woman with high-grade versus low-grade tumors demonstrated significantly higher risk of death (hazard ratio, 2.43; 95% confidence interval, 1.17-5.04; P = .02).
Ovarian cancer remains the most lethal gynecologic malignancy. Tumors of different histologies and grades have traditionally been grouped together and studied as a single disease entity. Recent developments, however, have questioned the wisdom of this strategy, especially as it applies to the treatment of women with serous carcinoma of the ovary.
Both histologic type and grade of tumors have been proven to be of prognostic significance for women with ovarian cancer. Prognostic significance of histologic grade has been complicated by the inclusion of all epithelial subtypes in the analyses within several studies.1-4 Nevertheless, no single system is universally used to grade these tumors. Our group has chosen to focus solely upon the grading of the most common histologic type, invasive serous carcinoma. We have developed and refined a simpler, clinically meaningful, and reproducible 2-tier grading system in which serous ovarian carcinomas are classified as low-grade or high-grade.5 This grading system is based primarily on the assessment of nuclear atypia, with mitotic rate used as a secondary feature (Figs. 1 and 2). Tumors with mild to moderate cytologic atypia are designated as low-grade, whereas tumors with marked cytologic atypia are designated as high-grade. Low-grade tumors tend to have a lower mitotic rate (usually up to 12 mitoses per 10 high-power fields [HPFs]). In contrast, high-grade tumors have conspicuous mitotic activity (>12 mitoses per 10 HPFs) and multinucleated cells.5 We have used this system at our institution for more than 15 years, and it has been subsequently validated and shown to have prognostic utility.6
Gershenson et al. described the clinical behavior of 112 patients with metastatic low-grade serous ovarian carcinoma.7 The data indicated that women with this type of tumor are younger and survive longer than women with high-grade tumors. Because the above study focused on patients from a single institution, we proposed to test the validity of the 2-tier grading system in a multi-institutional setting. We conducted an ancillary study of data from Gynecologic Oncology Group (GOG) Protocol 158, a prospective study of women with advanced epithelial ovarian cancer.8 The purpose of our study was 2-fold: 1) to use the 2-tier grading system to reclassify the histologic grade of tumors from women with serous ovarian carcinoma whose tumors had previously been classified using the International Federation of Gynecology and Obstetrics (FIGO) 3-tier system and 2) to determine the progression-free survival (PFS) and overall survival (OS) of patients treated on GOG 158 based upon reclassification of these tumors as low-grade or high-grade using the 2-tier system.9
MATERIALS AND METHODS
Before initiation of this study, institutional review board approval was obtained at The University of Texas MD Anderson Cancer Center. We retrospectively reviewed demographic, pathologic, and survival data of patients who participated in GOG Protocol 158, a cooperative multicenter prospective study of women with advanced epithelial ovarian cancer. Patients who underwent surgery with no residual mass >1 cm were randomized to postoperative cisplatin/paclitaxel (arm I) or carboplatin/paclitaxel (arm II). Our study focused on women with serous ovarian cancer treated with carboplatin/paclitaxel because of the labor-intensive pathology review and available resources. All patients signed an approved informed consent.
All pathology slides underwent central review by the GOG Pathology Committee for confirmation of eligibility. In the present study, 2 to 23 hematoxylin and eosin-stained slides of each primary ovarian tumor were reviewed by a gynecologic pathologist (M.T.D.) in a blinded fashion to confirm the original histologic diagnosis and independently reclassify the tumors using the 2-tier system. To strengthen the study, we expanded the group of reviewers by adding 5 additional pathologists (R.B., R.E., M.S., B.B., and G.R.), who also reviewed cases in an independent, blinded fashion. Despite time constraints, the vast majority of slides were reviewed by a minimum of 4 pathologists. The final 2-tier grade was determined by the grade agreed upon by the majority of reviewers. Pathologic data were correlated with clinical variables and survival. Patients were divided into 2 groups, low- and high-grade, for analysis according to tumor grade. The primary endpoints were PFS and OS. PFS was calculated from the date of study enrollment to the date of disease recurrence, death, or last contact. OS was calculated from the date of study enrollment to the date of death or last contact.
PFS and OS functions were estimated using the Kaplan-Meier procedure.10 The log-rank test was used to make comparisons based on tumor grade. A Cox proportional hazards model was used to confirm the independent prognostic value of the tumor grade adjusted for age, performance status (PS), and residual disease.10 These covariates were included for adjustment based on the prognostic variables identified by previous GOG studies.4 The Wilcoxon rank sums, Pearson chi-square, or the Fisher exact test was used to assess associations between tumor grade and patient clinicodemographic characteristics.11 Agreement between the 2-tier and FIGO grading systems was evaluated using kappa statistics.11 Comparisons were made between 2-tier low-grade versus FIGO grade 1 tumors and between 2-tier high-grade versus FIGO grade 2 and 3 tumors. All statistical tests were 2-tailed, with the significance level set at α = .05. Statistical analyses were performed using Statistical Analysis Software version 9.1 (SAS Institute, Cary, NC).
A total of 392 patients were enrolled on the carboplatin/paclitaxel arm of GOG Protocol 158. Of these, 290 patients were diagnosed by GOG pathologists as having serous carcinoma. Slides from 4 (1.4%) patients were not available. The tumors of 11 patients were excluded because of histologies other than serous ovarian cancer: 6 were borderline serous tumors, 2 were clear cell carcinoma, 2 were adenocarcinoma not otherwise specified, and 1 was mucinous carcinoma. To ensure the homogeneity of the study population, an additional 34 patients who did not complete the required chemotherapy were also excluded. Therefore, a total of 241 patients were included in the final analysis. Of the tumors of 241 patients, 15 were evaluated by 3 reviewers using the 2-tier system, 190 by 4 reviewers, and 36 by 5 reviewers. Therefore, the tumors of 226 (94%) patients were reviewed by a minimum of 4 pathologists. The reviewers showed high consistency in evaluation, with full agreement in the tumors of 197 (82%) patients.
Patient characteristics are summarized in Table 1. By using the 2-tier system, 21 patients (8.7%) were confirmed as having low-grade serous carcinoma, and 220 (91.3%) were confirmed as having high-grade disease. Table 2 shows associations of tumors graded by the 2-tier system with age, performance status, size of residual disease, and FIGO tumor grade.
Table 1. Patient Characteristics, N = 241
No. of Patients
Abbreviations: FIGO, Federation of Gynecology and Obstetrics; GOG, Gynecologic Oncology Group; MDACC, The University of Texas MD Anderson Cancer Center.
Table 2. Comparison of the Outcomes Associated With the 2-Tier Versus the FIGO Grading System
No. of Patients (%)
No. of Patients (%)
Abbreviations: FIGO, Federation of Gynecology and Obstetrics; MDACC, The University of Texas MD Anderson Cancer Center.
Associations were assessed using Wilcoxon test (for age), or Pearson chi-square or Fisher exact test (for categorical variables).
Median age, y
1 or 2
When FIGO grades 2 and 3 were combined, the overall agreement between the 2-tier and FIGO grading systems (low-grade vs FIGO grade 1 and high-grade vs FIGO grades 2 and 3) was 95% (230 of 241). The kappa statistic for this relationship was 0.68 (95% confidence interval [CI], 0.50-0.86), which was interpreted as substantial agreement. Of the 11 tumors graded differently by the 2-tier and FIGO grading systems, 7 tumors were originally characterized as FIGO grade 2 tumors, but were reclassified as low-grade in the 2-tier system; 3 tumors were originally classified as FIGO grade 1 but were reclassified as high-grade carcinomas.
With a median follow-up of 53 months, 206 (85.5%) patients progressed, and 183 (75.9%) died. By using the 2-tier system, patients with low-grade tumors had significantly longer median PFS compared with women with high-grade carcinomas (45.0 vs 19.8 months; P = .01); median OS for patients with high-grade carcinomas was 53.8 months, whereas median OS for patients with low-grade serous carcinomas was 126.2 months (P = .002; Table 3, Fig. 3). With the 2-tier system, 3-year PFS was 55.9% for patients with low-grade carcinomas and 27.7% for those with high-grade carcinomas. Furthermore, 3-year OS was 90.5% and 67.6%, respectively, for those with low-grade versus high-grade carcinomas (Table 3, Fig. 3). The independent prognostic value of the 2-tier system was further evaluated by multivariate analysis. Adjusted for age, performance status, and tumor residual, patients with tumors classified as high-grade using the 2-tier system were at increased risk for disease progression (hazard ratio [HR], 1.71; 95% CI, 0.98-2.99; P = .06) and also for death (HR, 2.43; 95% CI, 1.17-5.04; P = .02; Table 4).
Table 3. Estimates of PFS and OS According to the 2-Tier and FIGO Grading Systemsa
No. of Patients
3 Year, %
3 Year, %
Abbreviations: FIGO, Federation of Gynecology and Obstetrics; MDACC, The University of Texas MD Anderson Cancer Center; OS, overall survival; PFS, progression-free survival.
Survival was estimated by Kaplan-Meier procedure, and survival functions were compared using log-rank test.
Table 4. Comparison of Outcomes Associated With Tumor Grade Adjusted for Key Variablesa
Abbreviations: CI, confidence interval; FIGO, Federation of Gynecology and Obstetrics; HR, hazard ratio; MDACC, The University of Texas MD Anderson Cancer Center; OS, overall survival; PFS, progression-free survival.
HR estimated from Cox proportional hazards model adjusted for age group, performance status, and amount of tumor residual.
Clinical outcome using the FIGO grading system was also assessed. Median PFS for patients with FIGO grade 1, 2, and 3 tumors was 37.5, 19.8, and 20.1 months, respectively (P = .07). Median OS for FIGO grade 1 was 113.8 months, whereas median OS was 52.7 and 58.2 months for FIGO grade 2 and 3 tumors, respectively (P = .02; Table 3, Fig. 4). Three-year PFS was 50.0%, 23.1%, and 33.3%, respectively, for FIGO grade 1, 2, and 3 tumors. Three-year OS was 87.5% for patients with FIGO grade 1 tumors, 66.1% for those with grade 2 tumors, and 69.9% for those with grade 3 tumors. The adjusted results from the multivariate analysis were essentially the same (Table 4).
Generally, patients with FIGO grade 2 and 3 tumors had similar risk for poor prognosis. FIGO grade 2 and 3 tumors were associated with increased risk for disease progression and death when compared with FIGO grade 1 tumors. Importantly, median PFS, median OS, 3-year PFS, and 3-year OS were all shorter for patients with FIGO grade 2 tumors than for those with FIGO grade 3 tumors.
Our results demonstrate that serous carcinoma of the ovary can be effectively classified as 2 clinically distinct tumor types and that the 2-tier grading system, compared with the 3-tier FIGO grading system, offers a simpler yet precise framework for predicting clinical outcomes.
Despite the prognostic value associated with histologic type and tumor grade, no single grading system is universally used for serous carcinoma of the ovary, the most common histologic subtype of ovarian cancer. Multiple studies have demonstrated that evaluation of histology and grade is subjective and lacks both intraobserver and interobserver reliability.1, 12 Bertelsen and Anderson proposed 2 classification systems: 1 based on architectural pattern and another based on the combination of nuclear pleomorphism, mitotic count, and nucleocytoplasmic ratio.2 The rate of agreement between the 2 systems was only 62%.
Subsequently, Shimizu and Silverberg designed a schema based on architectural growth pattern, mitotic count, and nuclear grade.3 Predictions of survival according to the Shimizu-Silverberg system did not correlate well with those of the FIGO system, which is based primarily on the architectural structure of the tumor.3 Ishioka et al. further evaluated the utility of the Shimizu-Silverberg system by directly comparing it to the FIGO system as a prognostic indicator for ovarian cancer.13 Although the k value for the Shimizu-Silverberg grading system showed more agreement among pathologists than did the FIGO system, the majority of inconsistencies among the 3 independent pathologists using the Shimizu-Silverberg system were based on nuclear grade.
The above-described studies illustrate the importance of both histology and grade as prognostic indicators for women with epithelial ovarian cancer. Whereas significant work has been performed on a general level, our group has focused solely on invasive serous carcinoma of the ovary. We developed and refined a simple, clinically meaningful, and reproducible 2-tier grading system by classifying tumors as low-grade or high-grade. This system has been used at our institution for more than 15 years; tumors classified as low-grade using our system correspond extremely well with tumors classified as grade I by the Shimizu-Silverberg and FIGO systems, and the majority of high-grade tumors correspond to grade 2 and 3 tumors in those systems. Grade 2 and 3 tumors are grouped together because they both correspond with the high-grade serous carcinoma category.5 In addition, interobserver and intraobserver variability have been assessed among a group of 2 general surgical pathologists and 7 gynecologic pathologists using this system.6 Statistical analysis demonstrated an overall k statistic of 0.909 among the observers, indicating virtually perfect agreement.
Most recently, Seidman et al. reported on the comparison of a new binary system versus the 2-tier grading system for 113 cases of FIGO stage III serous carcinoma of the ovary and peritoneum.14 The authors hypothesized that a binary system would provide sufficient data for treatment decisions and would be more reproducible among pathologists. Although both systems are based on nuclear grade, the binary system has a higher threshold of nuclear size for diagnosing high-grade tumors.14 When Seidman and colleagues compared their new binary system with the 2-tier system, the authors concluded that the 2-tier grading system appears more promising, because it identifies a small but distinct group of serous tumors with a favorable prognosis.14
There is a significant and growing body of literature that provides molecular, genetic and clinical correlation with the 2-tier grading system and reinforces the existence of 2 distinct disease entities. DNA array studies by several groups demonstrated differential gene expression based upon a binary grading system.15-17 Studies focused upon BRAF, KRAS, and p53 mutations provide further support of unique genetic mutations among the 2 tumor grades.18-20 Recent molecular, genetic and in vitro drug resistance data also present additional evidence that subclassification of high-grade serous carcinoma into moderately-differentiated and poorly-differentiated categories is not relevant.21 On the basis of these data, Vang et al. suggested that grade 2 and grade 3 serous tumors simply be classified as high-grade serous carcinoma.21 Lastly, immunohistochemical comparisons between low-grade and high-grade carcinomas reveal significantly different levels of expression of a variety of proteins.22-24 Taken together, these data strongly suggest a unique underlying pathogenesis for low-grade and high-grade lesions.
With respect to clinical outcomes, Gershenson et al. indicated that women with low-grade serous carcinoma of the ovary are younger and have longer survival than women with high-grade disease.7 The long OS of 81.8 months is similar to that of patients with low-grade tumors described in the current study. Whereas the median PFS in the study by Gershenson et al. was less than that reported in the current study (19.9 and 42.5 months, respectively), the variance may be because of the smaller number of patients with low-grade disease in the current study and differences in chemotherapy regimens, disease stage, and amount of post-treatment residual disease reported in the 2 studies.7 Regardless of their distinctions, both of these studies strongly suggest different clinical courses for patients with low-grade and high-grade serous ovarian cancer.
Winter et al. recently combined data from 6 GOG protocols in an effort to identify factors predictive of poor prognosis in a similarly treated population of women with advanced epithelial ovarian cancer.4 The study sample included 1895 patients treated by primary surgery followed by platinum and taxane-based chemotherapy. Patient age, PS, tumor histology, and amount of residual tumor were identified as independent predictors of prognosis.4 Although tumor grade was not found to predict survival, tumors in this study were of all histologic subtypes. We believe that the wide variation in histology may have diluted the influence of tumor grade on PFS and OS. Because our study focused solely upon serous tumors and controlled for histology, we have evaluated the influence of tumor grade upon prognosis more rigorously.
Our data further reinforce the finding that distinguishing FIGO grade 2 tumors from FIGO grade 3 tumors is ambiguous at best. As shown in Table 4 and Figure 4, Top, the PFS event rate is lower among those classified as FIGO grade 3 when compared with FIGO grade 2. If the FIGO grading system were accurate, we would expect the opposite outcomes, that is, higher rates of disease progression and shorter survival in patients with FIGO grade 3 versus grade 2 disease. These seemingly counterintuitive observations strongly suggest that the FIGO grading system is not as clinically robust as we had once thought and should be further scrutinized.
The potential preference for use of the 2-tier grading system over the 3-tier system seems appropriate for several important reasons. First, the 2-tier system has minimal interobserver variability, demonstrated by the high level of agreement among the 6 participating pathologists. It is simple, user-friendly, and highly reproducible, characteristics specifically identified as crucial to the development of a universal grading system by all groups who have attempted to do so. Second, it appears to be clinically irrelevant to separate FIGO grade 2 and 3 tumors when grading serous ovarian carcinomas (Tables 3 and 4). Examination of the Kaplan-Meier survival curves, FIGO grade 2 and 3 survival curves and PFS point to poorer similar overall clinical outcomes in patients with FIGO grade 2 versus grade 3 tumors, even at the 5-year time period (Fig. 4). We are currently in an era in which novel therapeutics are being investigated, and potentially significant treatment advances may soon be based upon specific molecular and genetic tumor characteristics. Use of 2 categories rather than 3 to facilitate the study of low-grade serous carcinoma by triaging patients with low-grade serous carcinoma to appropriate clinical trials is not without possible errors. In the validation study, features that generated confusion in assignment of the grade in cases of low-grade serous carcinoma included high nuclear:cytoplasmic ratio, slightly irregular chromatin pattern, presence of nucleoli, and presence of cases with cribriform or solid architectural pattern. These histologic features are associated with less differentiated tumors at other sites and may have led to discrepant readings in the tumors graded differently by the 2-tier and FIGO grading systems.6 It is prudent to question the benefit of using the FIGO grading system in the face of repetitive inconsistencies and questionable clinical relevance.
Our study has several limitations. First, because no single grading system is infallible, reclassification using a 2-tier system is not without possible errors. An atypical clinical course should prompt the physician to question the accuracy of the pathology report. Second, one might argue that the results could be inaccurate because all 6 pathologists did not review every slide. However, as demonstrated previously, variability between observers is significantly less using the 2-tier grading system versus the 3-tier system. Although not perfect, the k statistic among the various observers was 0.909 in the validation study.6 To date, no studies have been published demonstrating such a high correlation among observers using the 3-tier grading system. Furthermore, the slides of 286 patients were re-reviewed in this study, yielding 9% low-grade tumors. Although larger numbers of low-grade tumors might be more reassuring, the shorter median PFS and OS in patients with FIGO grade 2 versus FIGO grade 3 tumors in our study, as well in the study by Winter et al., indicate that there is no clinical justification to separate these 2 grades.4 Lastly, although these findings apply to a population of women with advanced, optimally cytoreduced serous ovarian cancer, we believe that these results might also apply to patients with early-stage disease as well as those with advanced suboptimally cytoreduced disease. Further study is warranted.
In conclusion, when compared with the FIGO grading system, the 2-tier grading system for serous carcinoma of the ovary is a simple, yet precise, framework for predicting clinical outcomes. Use of the FIGO 3-tier grading system yields clinically ambiguous results. Our findings, bolstered by the information from previous clinical and molecular studies, clearly support the logic of separating women with serous ovarian carcinoma into 2 clinically distinct populations, low-grade and high-grade, in an effort to develop more effective therapies.4-7, 13-24 It is critical to acknowledge this clinicopathologic distinction as we move toward more personalized care of women with various subtypes of ovarian cancer. Utilization of the 2-tier system will ultimately improve the efficiency of clinical trials by triaging patients with rare tumors to trials solely evaluating those specific tumor types, thereby significantly impacting the standard of care for women with this disease. The separation of patients with low-grade and high-grade serous ovarian cancer is crucial, as our patients deserve unique consideration of and treatment for their specific histotypes.
This study was supported by National Cancer Institute grants to the Gynecologic Oncology Group Administrative Office (CA 27469) and Gynecologic Oncology Group Statistical and Data Center (CA 37517).