- Email: [email protected]
Ten-year relative survival for epithelial ovarian cancer
S34
ABSTRACTS / Gynecologic Oncology 120 (2011) S2–S133
Results: Two-hundred forty-four patients were identified who met H-IR criteria. One-hundred forty underwent surgery followed by observation, and 104 received adjuvant radiation therapy with or without chemotherapy. Recurrence rates for the two groups were similar at 11 and 8%. A univariate analysis was performed to assess the significance of individual predictors in the H-IR model as defined by H.M. Keys et al. (Gynecol Oncol 2004;92:744–51) in GOG 99. Both lymphovascular space invasion (LVSI) and invasion of the outer half of the myometrium were found to be significant predictors of recurrence (P = 0.003 and P = 0.05). On multivariate analysis, however, only LVSI remained significantly associated with disease recurrence. Kaplan–Meier curves were created for disease-specific survival in patients with and without adjuvant therapy. No difference in five-year disease-specific survival was observed between these cohorts on log-rank analysis (P = 0.39). Conclusions: In patients with surgically staged high-intermediate risk endometrial cancer, the overall risk of recurrence is low and is not significantly reduced when patients receive adjuvant therapy. Moreover, no survival benefit was demonstrated when adjuvant radiation therapy alone or in combination with chemotherapy was administered. Additionally, the use of adjuvant therapy in these generally low-risk patients does expose them to the risks of "multimodality treatment," which may increase overall complication rates. doi:10.1016/j.ygyno.2010.12.082
76 Retrospective review of an intraoperative algorithm to predict lymph node metastasis in low-grade endometrial adenocarcinoma P. Convery1, L. Cantrell2, S. Modesitt2, A. Alvarez-Secord1, L. Havrilesky1 1 Duke University Medical Center, Durham, NC, 2University of Virginia, Charlottesville, VA Objective: The role of lymphadenectomy (LND) is controversial in endometrial cancer (EMCA), and algorithms to identify women most likely to benefit from LND are required. We sought to validate the "Mayo" algorithm to identify women intraoperatively who do not require a lymphadenectomy. A multicenter retrospective chart review was completed using two independent institutional EMCA databases. Between 1977 and 2007, patients were identified with preoperative grade 1 or 2 EMCA. Inclusion criteria for the study were: (1) low-risk preoperative histology, excluding serous and clear cell; (2) no evidence of metastatic disease at surgery; (3) LND in which at least eight pelvic lymph nodes were sampled; (4) intraoperative pathology assessment. Patients were designated as satisfying the Mayo criteria if at the time of intraoperative assessment tumors were found to be grade 1 or 2 with endometrioid or endometrioid-like histology, myometrial invasion ≤ 50%, and tumor diameter ≤ 2 cm on intraoperative or final (if not measured intraoperatively) pathology. Nodal metastases and final staging were analyzed. Results: Of 442 patients meeting inclusion criteria, 110 satisfied the Mayo criteria. Two (1.8%) patients had positive lymph node metastases; final pathology in node-positive cases was stage IIIC1 papillary serous adenocarcinoma and stage IIIC1 endometrioid adenocarcinoma. The negative predictive value of the Mayo algorithm was 98.2%. One hundred four of 110 (94.5%) patients meeting the Mayo criteria were stage IA on final pathology using the 2009 FIGO staging criteria. Two patients each (1.8%) had stage IB, IIIA, and IIIC1 disease. Intraoperative analysis was consistent with final grade in 54 patients (49.1%), upgraded on final analysis in 13 patients (11.8%), and downgraded in two patients (1.8%). For 41 patients (37.3%), grade
was not available on either intraoperative or final pathology assessments because no invasive lesion was seen. Conclusions: At several institutions with more traditional pathology systems, the Mayo algorithm identified with a 98.2% negative predictive value women who would not benefit from a lymphadenectomy for endometrial cancer. Discrepancies between intraoperative and final pathology results may limit the general applicability of this algorithm. doi:10.1016/j.ygyno.2010.12.083
77 Identifying Lynch syndrome in women with endometrial cancer L. Chen, J. Holstein, A. Blanco, J. Chan, C. Powell, J. Rabban, J. Grenert, P. Conrad UCSF Helen Diller Family Comprehensive Cancer Center, San Francisco, CA Objective: Approximately 2% of endometrial cancers are attributable to Lynch syndrome. Selecting a subgroup of patients with endometrial cancer at higher risk for Lynch syndrome may identify a cohort for genetic counseling and testing. From January 2008 to June 2010, 328 women underwent hysterectomy for endometrial carcinoma. Mismatch repair protein (MMR) immunohistochemistry was prospectively performed for MLH1, MSH2, PMS2 and MSH6. Microsatellite instability (MSI) testing was performed by PCR and reported as microsatellite stable (MSS), low instability (MSI-low) or high instability (MSI-high). Inclusion criteria included: (1) age <50; (2) personal history of a Lynch cancer (colorectal, endometrial, urothelial, gastric, small bowel, brain, etc.); (3) a first-degree family member with a Lynch cancer, or pathology characteristics suggestive of Lynch cancer (tumor-infiltrating lymphocytes, peritumor lymphocytic infiltrate, mixed undifferentiated and well-differentiated carcinomas, lower uterine segment tumors, etc.). Results: One hundred twelve of 328 (34%) tumors fulfilled at least one criterion: 48 were under age 50, five had a personal history of colorectal cancer, 11 were diagnosed with a synchronous ovarian cancer, 36 had a notable family history, and 51 had suspicious pathologic features. All patients had MMR IHC performed on their tumor; 88 (79%) also had MSI testing. Thirty-one patients (28%) had loss of at least one or more MMR by IHC: 22 had loss of MLH1 and PMS2, one had loss of MSH2 and MSH6, four had loss of MSH6 only, and four had loss of PMS2 only. Among patients with loss of MMR by IHC, 26 of 31 (84%) underwent MSI testing and all but one were MSIhigh. Two patients with MSI-high tumors had intact MMR by IHC. All 31 patients with loss of MMR by IHC were contacted for genetic counseling; only 13 (42%) accepted. Nine of 13 patients underwent genetic testing, and two of the nine were found to have Lynch mutations (1 MLH1, 1 PMS2): one patient had a mother with uterine cancer, and the other was 42 years of age. Conclusions: The integrated use of clinical, histopathologic, and molecular markers may help identify women with endometrial cancer at risk for Lynch syndrome. Patient access and awareness of genetic counselors may help improve the identification process.
doi:10.1016/j.ygyno.2010.12.084
78 Ten-year relative survival for epithelial ovarian cancer L. Baldwin1, R. Ware1, B. Huang2, T. Tucker2, S. Goodrich1, I. Podzielinski1, C. DeSimone1, F. Ueland1, J. van Nagell1, L. Seamon1 1 University of Kentucky Medical Center, Lexington, KY, 2Kentucky Cancer Registry, Lexington, KY
ABSTRACTS / Gynecologic Oncology 120 (2011) S2–S133
Objective: The overall survival rate of patients with epithelial ovarian cancer (EOC) is poor and most of the patients who are alive at five years have active disease. Thus, 10-year survival may be a more appropriate endpoint. By definition, relative survival (RS) adjusts for the general survival rate of the population for that race, sex, age and date at which the diagnosis was coded. The 10-year RS for ovarian cancer using SEER data has previously been reported, but was not limited to epithelial histologies. Our objective was to measure RS in EOC over 10 years, stratified by stage, race, age, classification of residence and surgery as the first course of treatment. Using the SEER 1995–2007 database, cases were identified using the International Classification of Diseases codes for EOC. Inclusion criteria included malignant behavior, first primary, actively followed, and age ≥20 years. Exclusion criteria included death certificate- or autopsy-only cases. Using the actuarial life table method, RS over 10 years was calculated, stratified by stage, classification of residence, surgery as the first course of treatment, race and age. Results: Forty thousand six hundred ninety-two patients met inclusion criteria. Stage distribution was 20, eight, 39, 27, and 7% for stages I, II, III, and IV, and unknown, respectively. The first course of treatment was surgery in 78%. Overall RS was 65, 44, and 36% at two, five and 10 years, respectively. The slope of decline in RS was reduced for years five–10 as compared with years one–five after diagnosis. RS at five years was 89, 70, 36, and 17% for stages I, II, III, and IV, respectively. RS at 10 years was 84, 59, 23, and 8% for stages I, II, III, and IV, respectively. RS was comparable for patients living in rural and urban settings. At all stages, patients who underwent surgery as their first course of treatment had better RS than those who did not have surgery. When RS was compared between races, blacks had the poorest survival. At all stages, advanced age at time of diagnosis was associated with decreased RS. Conclusions: Although advanced EOC is associated with a poor prognosis, updated data demonstrate that survival rates adjusted for survival in the general population (RS) at five years are improved over historic reports and the 10-year RS for stage III is higher than expected. These data provide the physician and the patient with more accurate prognostic information, particularly for advanced-stage disease. doi:10.1016/j.ygyno.2010.12.085
79 Availability of gynecologic oncologists for ovarian cancer care D. Cooney1, B. Spruell1, S. Rim2, S. Foster2, S. Lawvere1, L. Richardson2, C. Thomas2, S. Stewart2 1 SciMetrika LLC, Durham, NC, 2Centers for Disease Control and Prevention, Atlanta, GA Objective: Women with ovarian cancer (OC) treated by gynecologic oncologists (GOs) receive guidelines-based treatment more often and have better clinical outcomes than those who are not. Disparities in OC treatment include some race- and age-specific populations receiving standard care less often. To assess whether geographic distribution of GOs contributes to these disparities, we used Geographic Information Systems (GIS) methods to describe a relationship between OC incidence rates and practicing GOs in the United States. Cancer incidence data were obtained from SEER and CDC's National Program of Cancer Registries. OC cases diagnosed from 2002 to 2006, covering 97.4% of the U.S. population, were ageadjusted using U.S. Census estimates (n = 106,391). County-level demographic, rural–urban and descriptive data were collected from census and area resource file data. A list of current GOs (n = 866) and
S35
their addresses was obtained from the Society of Gynecologic Oncologists; GO addresses were geocoded to determine county of location and distance measures from each county centroid to the nearest GO. Maps were generated using GIS to examine clustering and spatial relationships of GOs and OC incidence rates. Spatial analytic methods and correlation analyses were used to examine county-level characteristics. Multivariable logistic regression was conducted using dichotomous and polytomous categorization of incidence rates to determine the relationship between the availability of GOs and OC incidence. Results: The mean U.S. 5-year incidence rate of OC was 15.1 per 100,000 women in counties within 100 miles of a GO versus 18.0 for counties without a GO within 100 miles (P < 0.0001). There was a statistically significant (P < 0.0001) correlation between incidence rates and distance to a GO. Most GOs (99%) were located in the 1048 metropolitan counties, less than 1% were located in the 1271 urban counties, and none were located in the 597 rural counties. Conversely, OC incidence rates were higher in rural counties (19.2) than in urban (15.3) and metropolitan (14.1) counties (P < 0.0001). Conclusions: Gynecologic oncologists are unequally distributed in the United States, with the highest concentration in urban areas, where the ovarian cancer incidence rates are lowest. This inverse relationship may underlie existing treatment disparities for some groups of patients. Further exploration into additional barriers that affect availability and access to care for patients with ovarian cancer is needed. doi:10.1016/j.ygyno.2010.12.086
80 Biologic roles of tumor and endothelial delta-like ligand 4 in ovarian cancer W. Hu1, C. Lu1, R. Stone1, J. Bottsford-Miller1, A. Nick1, M. Shahzad2, K. Matsuo3, R. Coleman1, A. Sood1 1 University of Texas M.D. Anderson Cancer Center, Houston, TX, 2 University of Wisconsin, Madison, WI, 3University of Southern California, Los Angeles, CA Objective: The Notch/Delta-like ligand 4 (Dll4) pathway is emerging as an important target for new anti-angiogenesis approaches. The significance of Dll4/Notch signaling in ovarian cancer is not fully understood, and was examined in the current study. Dll4 expression was examined in 108 epithelial ovarian cancer samples (24 from patients with recurrent ovarian cancer treated with an anti-VEGF agent (aflibercept or bevacizumab)). The biologic importance of Dll4 was further investigated by in vitro and in vivo studies with Dll4 gene silencing. Results: Dll4 was overexpressed in 72% of tumors and was an independent predictor of poor survival. Among patients treated with aflibercept or bevacizuamb in combination with chemotherapy, the responders (complete response + partial response) had lower levels of Dll4 than those with stable or progressive disease (P < 0.001). In our orthotopic ovarian cancer model, VEGF increased Dll4 expression in tumor vasculature under hypoxic conditions. Immobilized Dll4 downregulated VEGFR2 expression in endothelial cells directly through methylation of the VEGFR2 promoter. Silencing Dll4 in tumor and tumor-associated endothelial cells resulted in inhibition of ovarian cancer growth and deregulation of angiogenesis (microvessel density and pericyte coverage), accompanied by induction of hypoxia in the tumor microenvironment. Furthermore, the combination of mouse and human Dll4 siRNA plus bevacizumab resulted in the greatest inhibition of tumor growth, reducing tumor size by 98 and 87% compared with controls, and by 93 and 46% compared with treatment with bevacizumab alone, in the SKOV3ip1 and A2780 murine ovarian cancer models, respectively.
ABSTRACTS / Gynecologic Oncology 120 (2011) S2–S133
Results: Two-hundred forty-four patients were identified who met H-IR criteria. One-hundred forty underwent surgery followed by observation, and 104 received adjuvant radiation therapy with or without chemotherapy. Recurrence rates for the two groups were similar at 11 and 8%. A univariate analysis was performed to assess the significance of individual predictors in the H-IR model as defined by H.M. Keys et al. (Gynecol Oncol 2004;92:744–51) in GOG 99. Both lymphovascular space invasion (LVSI) and invasion of the outer half of the myometrium were found to be significant predictors of recurrence (P = 0.003 and P = 0.05). On multivariate analysis, however, only LVSI remained significantly associated with disease recurrence. Kaplan–Meier curves were created for disease-specific survival in patients with and without adjuvant therapy. No difference in five-year disease-specific survival was observed between these cohorts on log-rank analysis (P = 0.39). Conclusions: In patients with surgically staged high-intermediate risk endometrial cancer, the overall risk of recurrence is low and is not significantly reduced when patients receive adjuvant therapy. Moreover, no survival benefit was demonstrated when adjuvant radiation therapy alone or in combination with chemotherapy was administered. Additionally, the use of adjuvant therapy in these generally low-risk patients does expose them to the risks of "multimodality treatment," which may increase overall complication rates. doi:10.1016/j.ygyno.2010.12.082
76 Retrospective review of an intraoperative algorithm to predict lymph node metastasis in low-grade endometrial adenocarcinoma P. Convery1, L. Cantrell2, S. Modesitt2, A. Alvarez-Secord1, L. Havrilesky1 1 Duke University Medical Center, Durham, NC, 2University of Virginia, Charlottesville, VA Objective: The role of lymphadenectomy (LND) is controversial in endometrial cancer (EMCA), and algorithms to identify women most likely to benefit from LND are required. We sought to validate the "Mayo" algorithm to identify women intraoperatively who do not require a lymphadenectomy. A multicenter retrospective chart review was completed using two independent institutional EMCA databases. Between 1977 and 2007, patients were identified with preoperative grade 1 or 2 EMCA. Inclusion criteria for the study were: (1) low-risk preoperative histology, excluding serous and clear cell; (2) no evidence of metastatic disease at surgery; (3) LND in which at least eight pelvic lymph nodes were sampled; (4) intraoperative pathology assessment. Patients were designated as satisfying the Mayo criteria if at the time of intraoperative assessment tumors were found to be grade 1 or 2 with endometrioid or endometrioid-like histology, myometrial invasion ≤ 50%, and tumor diameter ≤ 2 cm on intraoperative or final (if not measured intraoperatively) pathology. Nodal metastases and final staging were analyzed. Results: Of 442 patients meeting inclusion criteria, 110 satisfied the Mayo criteria. Two (1.8%) patients had positive lymph node metastases; final pathology in node-positive cases was stage IIIC1 papillary serous adenocarcinoma and stage IIIC1 endometrioid adenocarcinoma. The negative predictive value of the Mayo algorithm was 98.2%. One hundred four of 110 (94.5%) patients meeting the Mayo criteria were stage IA on final pathology using the 2009 FIGO staging criteria. Two patients each (1.8%) had stage IB, IIIA, and IIIC1 disease. Intraoperative analysis was consistent with final grade in 54 patients (49.1%), upgraded on final analysis in 13 patients (11.8%), and downgraded in two patients (1.8%). For 41 patients (37.3%), grade
was not available on either intraoperative or final pathology assessments because no invasive lesion was seen. Conclusions: At several institutions with more traditional pathology systems, the Mayo algorithm identified with a 98.2% negative predictive value women who would not benefit from a lymphadenectomy for endometrial cancer. Discrepancies between intraoperative and final pathology results may limit the general applicability of this algorithm. doi:10.1016/j.ygyno.2010.12.083
77 Identifying Lynch syndrome in women with endometrial cancer L. Chen, J. Holstein, A. Blanco, J. Chan, C. Powell, J. Rabban, J. Grenert, P. Conrad UCSF Helen Diller Family Comprehensive Cancer Center, San Francisco, CA Objective: Approximately 2% of endometrial cancers are attributable to Lynch syndrome. Selecting a subgroup of patients with endometrial cancer at higher risk for Lynch syndrome may identify a cohort for genetic counseling and testing. From January 2008 to June 2010, 328 women underwent hysterectomy for endometrial carcinoma. Mismatch repair protein (MMR) immunohistochemistry was prospectively performed for MLH1, MSH2, PMS2 and MSH6. Microsatellite instability (MSI) testing was performed by PCR and reported as microsatellite stable (MSS), low instability (MSI-low) or high instability (MSI-high). Inclusion criteria included: (1) age <50; (2) personal history of a Lynch cancer (colorectal, endometrial, urothelial, gastric, small bowel, brain, etc.); (3) a first-degree family member with a Lynch cancer, or pathology characteristics suggestive of Lynch cancer (tumor-infiltrating lymphocytes, peritumor lymphocytic infiltrate, mixed undifferentiated and well-differentiated carcinomas, lower uterine segment tumors, etc.). Results: One hundred twelve of 328 (34%) tumors fulfilled at least one criterion: 48 were under age 50, five had a personal history of colorectal cancer, 11 were diagnosed with a synchronous ovarian cancer, 36 had a notable family history, and 51 had suspicious pathologic features. All patients had MMR IHC performed on their tumor; 88 (79%) also had MSI testing. Thirty-one patients (28%) had loss of at least one or more MMR by IHC: 22 had loss of MLH1 and PMS2, one had loss of MSH2 and MSH6, four had loss of MSH6 only, and four had loss of PMS2 only. Among patients with loss of MMR by IHC, 26 of 31 (84%) underwent MSI testing and all but one were MSIhigh. Two patients with MSI-high tumors had intact MMR by IHC. All 31 patients with loss of MMR by IHC were contacted for genetic counseling; only 13 (42%) accepted. Nine of 13 patients underwent genetic testing, and two of the nine were found to have Lynch mutations (1 MLH1, 1 PMS2): one patient had a mother with uterine cancer, and the other was 42 years of age. Conclusions: The integrated use of clinical, histopathologic, and molecular markers may help identify women with endometrial cancer at risk for Lynch syndrome. Patient access and awareness of genetic counselors may help improve the identification process.
doi:10.1016/j.ygyno.2010.12.084
78 Ten-year relative survival for epithelial ovarian cancer L. Baldwin1, R. Ware1, B. Huang2, T. Tucker2, S. Goodrich1, I. Podzielinski1, C. DeSimone1, F. Ueland1, J. van Nagell1, L. Seamon1 1 University of Kentucky Medical Center, Lexington, KY, 2Kentucky Cancer Registry, Lexington, KY
ABSTRACTS / Gynecologic Oncology 120 (2011) S2–S133
Objective: The overall survival rate of patients with epithelial ovarian cancer (EOC) is poor and most of the patients who are alive at five years have active disease. Thus, 10-year survival may be a more appropriate endpoint. By definition, relative survival (RS) adjusts for the general survival rate of the population for that race, sex, age and date at which the diagnosis was coded. The 10-year RS for ovarian cancer using SEER data has previously been reported, but was not limited to epithelial histologies. Our objective was to measure RS in EOC over 10 years, stratified by stage, race, age, classification of residence and surgery as the first course of treatment. Using the SEER 1995–2007 database, cases were identified using the International Classification of Diseases codes for EOC. Inclusion criteria included malignant behavior, first primary, actively followed, and age ≥20 years. Exclusion criteria included death certificate- or autopsy-only cases. Using the actuarial life table method, RS over 10 years was calculated, stratified by stage, classification of residence, surgery as the first course of treatment, race and age. Results: Forty thousand six hundred ninety-two patients met inclusion criteria. Stage distribution was 20, eight, 39, 27, and 7% for stages I, II, III, and IV, and unknown, respectively. The first course of treatment was surgery in 78%. Overall RS was 65, 44, and 36% at two, five and 10 years, respectively. The slope of decline in RS was reduced for years five–10 as compared with years one–five after diagnosis. RS at five years was 89, 70, 36, and 17% for stages I, II, III, and IV, respectively. RS at 10 years was 84, 59, 23, and 8% for stages I, II, III, and IV, respectively. RS was comparable for patients living in rural and urban settings. At all stages, patients who underwent surgery as their first course of treatment had better RS than those who did not have surgery. When RS was compared between races, blacks had the poorest survival. At all stages, advanced age at time of diagnosis was associated with decreased RS. Conclusions: Although advanced EOC is associated with a poor prognosis, updated data demonstrate that survival rates adjusted for survival in the general population (RS) at five years are improved over historic reports and the 10-year RS for stage III is higher than expected. These data provide the physician and the patient with more accurate prognostic information, particularly for advanced-stage disease. doi:10.1016/j.ygyno.2010.12.085
79 Availability of gynecologic oncologists for ovarian cancer care D. Cooney1, B. Spruell1, S. Rim2, S. Foster2, S. Lawvere1, L. Richardson2, C. Thomas2, S. Stewart2 1 SciMetrika LLC, Durham, NC, 2Centers for Disease Control and Prevention, Atlanta, GA Objective: Women with ovarian cancer (OC) treated by gynecologic oncologists (GOs) receive guidelines-based treatment more often and have better clinical outcomes than those who are not. Disparities in OC treatment include some race- and age-specific populations receiving standard care less often. To assess whether geographic distribution of GOs contributes to these disparities, we used Geographic Information Systems (GIS) methods to describe a relationship between OC incidence rates and practicing GOs in the United States. Cancer incidence data were obtained from SEER and CDC's National Program of Cancer Registries. OC cases diagnosed from 2002 to 2006, covering 97.4% of the U.S. population, were ageadjusted using U.S. Census estimates (n = 106,391). County-level demographic, rural–urban and descriptive data were collected from census and area resource file data. A list of current GOs (n = 866) and
S35
their addresses was obtained from the Society of Gynecologic Oncologists; GO addresses were geocoded to determine county of location and distance measures from each county centroid to the nearest GO. Maps were generated using GIS to examine clustering and spatial relationships of GOs and OC incidence rates. Spatial analytic methods and correlation analyses were used to examine county-level characteristics. Multivariable logistic regression was conducted using dichotomous and polytomous categorization of incidence rates to determine the relationship between the availability of GOs and OC incidence. Results: The mean U.S. 5-year incidence rate of OC was 15.1 per 100,000 women in counties within 100 miles of a GO versus 18.0 for counties without a GO within 100 miles (P < 0.0001). There was a statistically significant (P < 0.0001) correlation between incidence rates and distance to a GO. Most GOs (99%) were located in the 1048 metropolitan counties, less than 1% were located in the 1271 urban counties, and none were located in the 597 rural counties. Conversely, OC incidence rates were higher in rural counties (19.2) than in urban (15.3) and metropolitan (14.1) counties (P < 0.0001). Conclusions: Gynecologic oncologists are unequally distributed in the United States, with the highest concentration in urban areas, where the ovarian cancer incidence rates are lowest. This inverse relationship may underlie existing treatment disparities for some groups of patients. Further exploration into additional barriers that affect availability and access to care for patients with ovarian cancer is needed. doi:10.1016/j.ygyno.2010.12.086
80 Biologic roles of tumor and endothelial delta-like ligand 4 in ovarian cancer W. Hu1, C. Lu1, R. Stone1, J. Bottsford-Miller1, A. Nick1, M. Shahzad2, K. Matsuo3, R. Coleman1, A. Sood1 1 University of Texas M.D. Anderson Cancer Center, Houston, TX, 2 University of Wisconsin, Madison, WI, 3University of Southern California, Los Angeles, CA Objective: The Notch/Delta-like ligand 4 (Dll4) pathway is emerging as an important target for new anti-angiogenesis approaches. The significance of Dll4/Notch signaling in ovarian cancer is not fully understood, and was examined in the current study. Dll4 expression was examined in 108 epithelial ovarian cancer samples (24 from patients with recurrent ovarian cancer treated with an anti-VEGF agent (aflibercept or bevacizumab)). The biologic importance of Dll4 was further investigated by in vitro and in vivo studies with Dll4 gene silencing. Results: Dll4 was overexpressed in 72% of tumors and was an independent predictor of poor survival. Among patients treated with aflibercept or bevacizuamb in combination with chemotherapy, the responders (complete response + partial response) had lower levels of Dll4 than those with stable or progressive disease (P < 0.001). In our orthotopic ovarian cancer model, VEGF increased Dll4 expression in tumor vasculature under hypoxic conditions. Immobilized Dll4 downregulated VEGFR2 expression in endothelial cells directly through methylation of the VEGFR2 promoter. Silencing Dll4 in tumor and tumor-associated endothelial cells resulted in inhibition of ovarian cancer growth and deregulation of angiogenesis (microvessel density and pericyte coverage), accompanied by induction of hypoxia in the tumor microenvironment. Furthermore, the combination of mouse and human Dll4 siRNA plus bevacizumab resulted in the greatest inhibition of tumor growth, reducing tumor size by 98 and 87% compared with controls, and by 93 and 46% compared with treatment with bevacizumab alone, in the SKOV3ip1 and A2780 murine ovarian cancer models, respectively.