Survivorship in advanced ovarian cancer: a prognostic model for overall survival and risk of recurrence

Abstract

Objective

Advanced epithelial ovarian cancer (EOC) poses a significant clinical challenge due to its typically late diagnosis and poor prognosis. However, a subset of patients exhibit remarkably prolonged survival. Identifying prognostic factors and developing tools for estimating outcomes may provide tailored strategies for treatment escalation or de-escalation. This study aimed to identify prognostic factors associated with patient survival and develop a prognostic model estimating EOC patients’ overall survival and risk of recurrence (ROR).

Methods

We conducted a retrospective analysis of 1049 women diagnosed with EOC from January 2002 until June 2024. Clinical, pathological, and molecular data, including germline BRCA pathogenic variants (PVs), and homologous recombination repair analysis were performed. Long-term survivors (LTS), defined as those surviving over 7 or 10 years, and short-term survivors (STS), defined as those surviving less than 2 years were compared. A prognostic model was developed using multivariable logistic regression to estimate survival probabilities and recurrence risk.

Results

Among the study cohort with advanced disease (FIGO stage III-IV), 20.3% survived beyond 7 years and 9.8% beyond 10 years. Factors significantly associated with LTS included younger age, lower disease stage, complete tumor resection, BRCA PV, and treatment with poly (ADP-ribose) polymerase inhibitors. The prognostic model, integrating age, stage, BRCA status, and tumor resection, provided survival estimates and ROR for 2, 5, 7, and 10 years from diagnosis. This tool is based on retrospective logistic regression analysis of long-term and STS across all stages (I-IV).

Conclusions

This study reaffirms established prognostic factors of LTS with advanced EOC and introduces a novel prognostic calculator integrating clinical variables. The tool may assist in personalizing treatment plans and guiding clinical decisions. Validation in multi-institutional cohorts is necessary to confirm its universal utility and applicability.

Implications for Practice.

Advanced epithelial ovarian cancer (EOC) patients have a poor prognosis; however, some patients achieve remarkably prolonged survival. This study aimed to identify prognostic factors and develop a model to estimate survival and risk of recurrence. A retrospective analysis of 1049 EOC patients diagnosed between 2002 and 2024 included clinical, pathological, and molecular data, comparing long-term survivors ( >7 or >10 years) and short-term survivors (<2 years). Long-term survivors were younger, had lower disease stage, complete tumor resection, presence of germline pathogenic variants in BRCA, and treatment with poly (ADP-ribose) polymerase inhibitors. A novel prognostic tool using multivariable logistic regression was developed to estimate survival and recurrence risks at 2, 5, 7, and 10 years. The model incorporates variables such as age, stage, BRCA status, and tumor resection. The prognostic calculator introduces a personalized approach to EOC management, potentially guiding tailored treatment strategies.

Introduction

Ovarian cancer is a leading cause of death among gynecologic malignancies, presenting a major clinical challenge due to its frequent diagnosis at an advanced stage.¹ In over two-thirds of cases, this delay is attributed to nonspecific symptoms.² The standard treatment for patients diagnosed with advanced epithelial ovarian cancer (EOC) involves cytoreductive surgery and platinum-based chemotherapy, often complemented by maintenance therapy with bevacizumab and poly (ADP-ribose) polymerase (PARP) inhibitors in select patients with relatively high response.³ Despite these therapeutic advances, the prognosis for-stage III-IV EOC remains poor, with less than half of patients surviving beyond 5 years.²

A small subset of women with advanced EOC attain exceptionally long-term survival, while others exhibit resistance to therapy, rapid disease progression, and ultimately succumb to their disease.^4–7

Prior studies have identified several favorable prognostic factors associated with long-term survivors (LTS). These include younger age at diagnosis, early-stage disease, non-serous histology, optimal cytoreduction, platinum sensitivity, and long recurrence-free survival.^8–10 Additional prognostic measures of extended survival include the absence of ascites and lower premaintenance CA-125 levels.^4,11 However, some LTS present with traditionally poor prognostic features such as suboptimal cytoreduction or platinum-resistant disease, suggesting that other yet unrecognized factors may contribute to their favorable outcomes.^12,13

Molecular characteristics also appear to influence survival. Women who survive longer more frequently harbor germline or somatic pathogenic variants (PVs) in BRCA1/2, exhibit homologous recombination deficiency (HRD), an alteration in the homologous recombination DNA repair pathway, and display immune-related characteristics such as higher levels of CD8⁺ tumor-infiltrating lymphocytes and increased Ki-67 staining.^5,14

Given the heterogeneity of clinical outcomes in EOC, this study aimed to characterize patients with exceptionally favorable or poor survival and identify clinical, pathological, and molecular features associated with long- and short-term survival outcomes. Our objective was to develop a prognostic calculator that integrates key prognostic variables into a weighted model to predict overall survival (OS) and risk of recurrence (ROR), ultimately supporting more personalized treatment strategies for women with advanced EOC.

Methods

Study population

We conducted a retrospective longitudinal analysis of 1049 women diagnosed and treated with EOC at Tel Aviv Sourasky Medical Center (TASMC), Israel, from January 2002 to June 2024. The study was conducted in accordance with the Declaration of Helsinki and approved by the Institutional Ethics Committee of TASMC (Identifier: 0346-20-TLV). The study cohort included women over 18 years of age diagnosed with FIGO stage III-IV EOC of various histologies, including serous-papillary, endometrioid, and other histology (clear cell, carcinosarcoma, and mucinous ovarian cancer). Patients with non-epithelial histology, low-grade serous or borderline histology were excluded from the study. Women with stage I-II disease were excluded from the analysis of survivorship, as women with early-stage disease are offered upfront surgery and have a comparatively good prognosis.

Data collection

Patient data were extracted from medical charts and included the following variables: demographics, past medical history, age, stage of disease at diagnosis, dates of diagnosis and surgery, optimal/suboptimal cytoreductive surgery (complete resection [R0], residual tumor ≤1 cm [R1], or residual tumor >1 cm [R2]), germline and somatic BRCA PV status, mean and median Cancer Antigen 125 (CA125) tumor marker levels at diagnosis, after 4 and 6 cycles of chemotherapy, the CA125 kinetics-the rate of CA125 decline, chemotherapy scheduling regimens (paclitaxel-carboplatin every 3 weeks [PC-3W] and weekly [PC-W]), and administration of maintenance PARP inhibitors and bevacizumab.

Genetic and genomic evaluation

Genetic testing was performed to identify germline PV in BRCA1/2 as well as somatic alterations in BRCA1/2, alongside molecular profiling for HRD. Homologous recombination deficiency testing became available in Israel after 2021, when it was approved for reimbursement by the Israeli Ministry of Health. The testing followed ministry guidelines, which stipulated that eligible patients are those who tested negative for common germline BRCA1/2 PV. Consequently, the cohort of HRD-positive patients included individuals with homologous recombination defects in DNA repair pathways and somatic BRCA mutations, while excluding those harboring germline BRCA1/2 PV. Following approval from the Ministry of Health, patients in our cohort who tested HRD-positive or harbored a BRCA PV received PARP inhibitor maintenance therapy either after completing first-line platinum-based chemotherapy or at recurrence, provided they had responded to platinum-based treatment.

Outcomes

Long term survivors were defined as women with advanced EOC (FIGO stage III-IV) who had survived more than 7 or 10 years from diagnosis, and short-term survivors (STS) were defined as women with advanced EOC who survived less than 2 years from diagnosis.

Study outcomes, including progression-free survival (PFS) and OS, were assessed across various patient populations characterized by genetic and molecular profiles. Progression-free survival was calculated from the first treatment to either progression or death or to the last known follow-up. Overall survival was calculated from the diagnosis date to either death or to the last known follow-up. Patient subpopulations included those with a germline BRCA1/2 PV, BRCA wild-type (WT), individuals with HRD, and those with homologous recombination proficient (HRP) disease.

Additionally, we aimed to develop a risk calculator to predict the likelihood of long-term or short-term survival based on the identified prognostic factors in the study population.

Statistical analysis

Statistical analysis was performed using R version 4.4.1 (R Development Core Team). Continuous variables were summarized as median and range and compared across groups by t-test. Categorical variables were summarized as a number and percentage and compared across groups by chi-square test. Survival functions were demonstrated using the Kaplan-Meier method. Survival distributions were compared across groups using the Log-Rank test. Cox proportional hazards regression was used to assess the effect of age stage, BRCA PV, CA125, histological subtype, suboptimal/optimal debulking, PARP inhibitor, bevacizumab on survival outcomes. Results were presented as hazard ratio (HR) with a 95% CI. Multivariable backward stepwise logistic regression analysis was performed to predict the effect of clinical variables on survival. The rate of CA125 decline was measured by the slope of the linear regression. P-values <.05 were considered statistically significant.

Bevacizumab in Israel is typically reserved for patients with high-risk disease, specifically those with FIGO stage IV or stage III disease with residual tumor following cytoreductive surgery (R1, R2). To account for this selective use and minimize confounding by indication, we performed a subgroup analysis restricted to this high-risk population. Multivariate logistic regression models were used to compare STS (<2 years) with LTS (>5, >7, and >10 years) in this subgroup. Covariates in the model included age at diagnosis, FIGO stage (III-IV), BRCA PV status, histologic subtype (serous vs non-serous), suboptimal/optimal debulking, neoadjuvant chemotherapy (NAC) followed by interval debulking, and bevacizumab administration. This approach enabled a more accurate assessment of the association between bevacizumab treatment and survival outcomes by isolating the analysis to patients for whom the treatment was clinically indicated.

Moreover, to ensure adequate follow-up for evaluating long-term survival, we analyzed a separate subset of patients diagnosed before 2017, allowing for a minimum of 7 years of follow-up.

Risk calculator

With the identification of key prognostic parameters among women with advanced disease (stages III-IV), we developed a series of prognostic risk assessment calculators specifically for all EOC stages I-IV. These calculators incorporate 4 key variables: age, stage of disease, tumor resection, and BRCA status. Based on our retrospective logistic regression analysis, they provide probabilities of survival and recurrence at different time points (2, 5, 7, and 10 years from diagnosis). Patients who survived more than 10 years were categorized to include those who also survived more than 2, 5, and 7 years. Similarly, patients who survived more than 7 years included those who also survived more than 5 and 2 years. The calculation involves creating a linear combination of variables by summing the intercept (baseline value) and the weighted contributions of each parameter. This linear combination is then converted into a probability using the sigmoid function, with the resulting probability expressed as a percentage. Each calculator is tailored to specific survival segments or events, such as the likelihood of surviving beyond 2, 5, 7, or 10 years, or experiencing recurrence within these timeframes, offering estimated risk assessments.

Results

Patient population

The study comprised 1049 women diagnosed with EOC and treated at a single center from January 2002 to June 2024. Nine hundred ten (87%) patients were diagnosed with the International Federation of Gynecology and Obstetrics (FIGO) stage III-IV disease, while a smaller patient population was diagnosed with stage I-II (139,13%). The evaluation of prognostic parameters included patients with advanced disease who received either primary cytoreduction followed by adjuvant chemotherapy or NAC followed by interval cytoreduction, excluding patients with early-stage disease (I-II). Less than half of the patients underwent primary debulking surgery (n = 387, 42.5%), and the remainder were referred to interval debulking surgery (n = 523, 57.5%).

Clinical characteristics of LTS- and STS

Among women with advanced disease, 184 patients (20.3%) survived >7 years after EOC diagnosis, 89 patients (9.8%) survived >10 years, and 143 (15.7%) survived <2 years from diagnosis (Table 1). The median age of LTS was significantly lower compared to the median age of STS (58 years [range 33-85 for LTS >7 years and 33-79 for LTS >10 years] vs 70 years for STS [range 24-91], P < .0001). Most LTS were diagnosed with stage III (94.02% for LTS >7 years and 96.63% for LTS >10 years) compared to STS <2 years (68.53%, P < .0001). More than half of the LTS harbored a BRCA PV (50.99% for LTS >7 years and 57.75% for LTS >10 years), compared to only 12.79% of STS <2 years (P < .0001).

Table 1.

Comparison of the clinical characteristics of long and short-term survivors with advanced EOC.

Variable	Total population n = 910	Survived <2 years N = 143 (15.7%)	Survived >7 years N = 184 (20.3%)	P-value	Survived >10 years N = 89 (9.8%)	P-value
Diagnosis age years, median (range)	62 (26-93)	70 (24-91)	58 (33-85)	<.0001	58 (33-79)	<.0001
Stage, n (%)				<.0001		<.0001
Stage III	741/1049 (70.64%)	98/143 (68.53%)	173/184 (94.02%)		86/89 (96.63%)
Stage IV	169/1049 (16.11%)	45/143 (31.47%)	11/184 (5.98%)		3/89 (3.37%)
Histology				.4364		.5066
Serous papillary	725/1051 (68.98%)	84/143 (58.74%)	118/184 (64.13%)		59/89 (66.29%)
Endometrioid	291/1051 (27.69%)	52/143 (36.36%)	61/184 (33.15%)		26/89 (29.21%)
Other	35/1051 (3.33%)	7/143 (4.9%)	5/184 (2.72%)		4/89 (4.49%)
BRCA status, n (%)						<.0001
BRCA 1	196/773 (25.36%)	8/86 (9.3%)	53/151 (35.1%)		25/71 (35.21%)
BRCA 2	81/773 (10.48%)	3/86 (3.49%)	24/151 (15.89%)	<.0001	16/71 (22.54%)
Negative	496/773 (64.17%)	75/86 (87.21%)	74/151 (49.01%)		30/71 (42.25%)
Unknown	268/1041 (25.74%)	57/143 (39.86%)	31/182 (17.03%)		18/89 (20.22%)
Residual disease, n (%)				<.0001		<.0001
R0		51 (36.69%)	147 (82.12%)		78 (88.64%)
R1/R2		88 (63.31%)	32 (17.88%)	<.0001	10 (11.36%)	<.0001
Time to recurrence months, median (range)	15.24 (0-267.33)	2.89 (0-18.4)	48.13 (4.99-233.99)		93.7 (4.99-233.99)

Most LTS had complete tumor resection (R0) compared to STS (82.12% for LTS >7 years and 88.64% for LTS >10 years vs 36.69% for STS <2 years, P < .0001).

No significant differences in histology, pretreatment CA125 levels, or CA125 values after 4 and 6 cycles of chemotherapy were observed between LTS and STS. However, an exploratory analysis of CA125 kinetics among patients with stage III and IV disease revealed a steeper decline in CA125 slope among those with longer survival. A significant difference was observed between STS (<2 years) and LTS (>5 years) (P = .04), with a trend toward significance between STS (<2 years) and LTS (>7 years) (P = .057). Comparisons involving LTS >10-year survivors did not reach statistical significance (STS <2 years vs. LTS >10 years: P = .18).

The proportion of patients who received bevacizumab as maintenance treatment was not statistically significantly different between LTS and STS (17% (n = 42) of LTS > 7 years and 16.85% (n = 15) of LTS > 10 vs 20.98% (n = 30) of STS < 2 years).

In contrast, 20.65% (n = 51) of LTS > 7 years and 26.97% (n = 24) of LTS > 10 received PARP inhibitors as maintenance treatment, while none of the STS < 2 received PARP inhibitors as maintenance therapy (P < .0001). The first PFS was significantly longer among the LTS (48.13 months [95% CI, 4.99-233.99] for >7 years and 93.7 months [95% CI, 4.99-233.99] for >10 years) compared to STS (2.89 months [95% CI, 0-18.40], P < .0001). Most LTS had a platinum-sensitive disease (98.91% of LTS >7 years and 98.88% LTS >10 years) compared to 24.82% of STS <2 years (P < .0001).

Prognostic measures of LTS

Multivariate logistic regression models comparing STS (<2 years) with LTS (>5, >7, and >10 years) showed that younger age at diagnosis, the presence of germline or somatic BRCA PV, and optimal cytoreduction (R0) were significantly associated with LTS across all timepoints (Table S1). BRCA PV conferred a particularly strong survival advantage, with odds ratios (ORs) increasing from 6.9 for >5-year survival to 11.8 for >10-year survival. Conversely, more advanced-stage disease (stage IV vs stage III) and suboptimal debulking (R1, R2) were strongly associated with early mortality. Neoadjuvant chemotherapy followed by interval cytoreduction was inversely associated with long-term survival. Histological subtype (serous vs non-serous) was not a significant prognostic measure.

LTS with extended follow-up

Multivariable logistic regression analysis was used to evaluate survival in patients who survived at least 7 years from diagnosis. We analyzed a cohort of patients diagnosed before 2017, ensuring a follow-up period of over 7 years to minimize bias in long-term outcome assessment (N = 559).

Factors associated with extended survival included harboring a BRCA PV (OR = 1.76, P = .02), treatment with PARP inhibitors (OR = 2.97, P = .0007), complete tumor resection, (OR = 0.256, P < .0001), younger age (OR = 0.98, P = .03), and earlier stage of disease (OR = 0.39, P = .02). These findings reaffirmed the previous analysis. The debulking approach was not significantly associated with survival beyond 7 years (OR = 0.67, P = .05), nor was bevacizumab therapy (OR = 0.70, P = .14).

Bevacizumab in high-risk patients

To account for the selective administration of bevacizumab in high-risk patients, we conducted a subgroup analysis limited to patients with stage IV or stage III disease with suboptimal resection and residual tumor (R1/R2). In multivariate logistic regression models comparing STS (<2 years) with LTS (>5, >7, and >10 years), bevacizumab use was not significantly associated with extended survival at any time point (Table S2). Odds ratios ranged from 1.10 to 1.65 across models, with wide confidence intervals and nonsignificant P-values (P > .13 in all models), suggesting no clear survival advantage. Notably, younger age and presence of a BRCA PV remained significant prognostic measures of LTS, while stage and histology were not significant.

Genetic, molecular profile, and survivorship

Survival outcomes varied among the patients based on their genetic and molecular profiles. Patients with germline BRCA PV showed significantly longer survival compared to patients who were BRCA WT, with a median OS of 82.1 months (95% CI, 73.59-105.53) vs 50.5 months (95% CI, 46.52-55.39) (P < .0001) and a median PFS of 19.58 months (95% CI, 17.71-23.00) vs 12.62 months (95% CI, 11.47-14.52) (P = .003, Table S3).

Within the cohort with BRCA PV, those with BRCA2 mutations had an extended survival compared to patients with BRCA1 mutations. Specifically, the median OS was 112.03 months (95% CI, 81.35-NA) for BRCA2 vs 74.35 months (95% CI, 69.22-97.68) for BRCA1 (P = .043), and the median PFS was 30.65 months (95% CI, 19.71-39.56) for BRCA2 compared to 18 months (95% CI, 15.90-21.03) for BRCA1 (P = .013).

Among the 221 patients who underwent HR testing, 115 (52%) were classified as HRP, 89 (40.3%) HRD, and 17 (7.7%) had inconclusive results. Patients with an HRD profile had a longer survival compared to HRP patients, with a median OS of 109.44 months (95% CI, 93.27-NA) vs 63.24 months (95% CI, 48.16-NA) (P = .002) and a median PFS of 17.71 months (95% CI, 13.90-34.60) vs 12.45 months (95% CI, 11.33-18.86) (P = .057).

BRCA mutation type and survival outcomes

To further evaluate the prognostic significance of BRCA PV, we conducted additional analyses stratified by the affected gene (BRCA1 vs BRCA2) and by specific mutation type. Long-term survivors (>5, >7, and >10 years) with BRCA1/2 PV were analyzed based on mutation location, with particular attention to functional domains of BRCA1 (Really Interesting Gene, DNA-binding domain [DBD], or C-terminal domain of BRCA1 [BRCT]) and BRCA2 (RAD51-binding domain [RAD51-BD]; DBD). Across all survival thresholds (>5, >7, and >10 years), the distribution of specific BRCA mutations (6174delT, 5382insC, 185delAG, and other mutations) and their locations within functional domains were not significantly associated with LTS. Fisher’s exact test P-values for mutation-type comparisons were >0.05 for all survival cutoffs (Table S4A and B). Similarly, no significant differences in survival outcomes were observed between BRCA1 and BRCA2 mutation carriers.

Prognostic model calculator for survival and risk of recurrence

We developed a prognostic model to estimate OS and ROR. The calculator integrated 4 key prognostic parameters: age, stage, BRCA status, and extent of tumor resection. This tool is based on retrospective logistic regression analysis of long-term and STS across all stages (I-IV). The calculator estimates the likelihood of survival beyond 2, 5, 7, or 10 years, as well as the probability of disease recurrence within these time frames.

For example, a 50-year-old patient with stage III disease, with a PV in BRCA, and incomplete cytoreductive surgery (R1 or R2) has a 94.6% chance of surviving 2 years. This patient population has a 37% chance of surviving 7 years and a 15.4% chance of surviving 10 years. In comparison, a 70-year-old patient with similar characteristics has a 25% chance of surviving beyond 7 years and a 13% chance of surviving 10 years, as demonstrated in Figure 1, describing the likelihood of surviving 7 years from diagnosis. The prognostic models for OS and ROR at 2,5,7, and 10 years are presented in Figure S1.

Figure 1.

Equation estimating the probability (%) of surviving beyond 7 years from diagnosis.

Figure 1 and Figure S1 demonstrate the survival calculator based on a logistic regression model used to estimate the chance that a patient will survive beyond a specific time point (>2, >5, >7, >10 years), based on clinical and molecular characteristics.

Each characteristic, such as age, stage at diagnosis, BRCA, and extent of tumor resection, is assigned a numerical weight that reflects how strongly it influences survival. These weights are based on multivariate analysis of outcomes in a large patient cohort. For example, younger age and the presence of a BRCA PV contribute positively to survival, while advanced stage or residual disease after surgery has a negative impact.

To generate a survival estimate for an individual patient, the model calculates a score by multiplying each feature by its respective weight and summing the results. This score represents the patient’s overall risk profile. The score is then transformed into a percentage using the logistic sigmoid function: Probability (in %) = 100 × [e^(score)/(1 + e^(score))]

This formula ensures that the result is always between 0 and 100 and represents the estimated chance of surviving beyond the chosen timepoint. The output provides an individualized estimate that supports clinical decision-making based on routinely available patient data.

Discussion

This real-world analysis compared LTS- and STS to identify key prognostic factors in women with advanced EOC. Consistent with established literature, our work reaffirms that younger age, lower stage, optimal cytoreduction, BRCA PV, and HRD molecular profile are positive prognostic variables in shaping survival trajectories in advanced EOC.^{4,  15,16} The use of NAC and interval cytoreduction was inversely associated with LTS, possibly reflecting its more frequent application in higher-risk patients. Previous studies have suggested that the survival effect of NAC for advanced EOC may differ based on patient and tumor factors.^17–19 Such that in older women, stage IV disease, and greater disease extent, NAC was associated with similar OS compared to primary debulking surgery.^20,21

With the identification of prognostic parameters of survivorship, we developed a prognostic risk calculator incorporating 4 key variables: age, BRCA status, tumor resection, and stage of disease at diagnosis. The calculator, developed from retrospective logistic regression analysis of both LTS- and STS, estimates the likelihood of surviving beyond 2, 5, 7, or 10 years and ROR at these time points. This tool can assist clinicians in tailoring treatment strategies, potentially guiding decisions on intensification or de-escalation of therapy.

The survival analysis based on genetic and molecular profiles further highlighted the heterogeneity in ovarian cancer outcomes. Patients with BRCA PV, particularly those harboring BRCA2 PV, demonstrated significantly extended OS and PFS compared to BRCA-WT counterparts. Moreover, patients with an HRD molecular profile exhibited prolonged survival comparable to those with BRCA2 mutations. This work underscores the prognostic and therapeutic importance of genetic and genomic testing, as BRCA PV and HRD tumors respond well to platinum-based chemotherapy and PARP inhibitors, validating findings from existing literature in a real-world, single-center patient population.^22–25

Although the presence of a BRCA PV was strongly linked with LTS, our results suggest that among BRCA PV carriers, neither the specific mutation nor the functional domain affected (BRCA1/2) was associated with differential survival outcomes in the LTS (>5, >7, and >10 years). The observed survival advantage in advanced-stage EOC among BRCA PV carriers appears consistent irrespective of mutation type. The absence of significant associations may, in part, reflect the relatively limited number of patients within individual mutation subgroups, reducing statistical power. These findings are consistent with previous literature; notably, a subgroup exploratory analysis from the PAOLA-1/ENGOT-ov25 trial demonstrated that maintenance olaparib provided a PFS benefit in patients with advanced-stage BRCA-mutated EOC, regardless of mutation location, when administered alongside bevacizumab.²⁶

Interestingly, the multivariable logistic regression analysis of LTS with advanced EOC and extended follow-up revealed that complete tumor resection was the strongest factor associated with extended survival, followed by BRCA mutational status, PARP-inhibitors treatment, younger age, and earlier stage at diagnosis. In contrast, the debulking approach and treatment with bevacizumab were not significantly associated with survivorship over 7 years.

Our study demonstrated the limited prognostic role of the blood marker CA125. Elevated CA125 levels at diagnosis did not significantly correlate with survivorship when comparing LTS to STS. Elevated CA125 may sometimes be attributable to ascites and normalize with therapy. However, exploratory analysis of CA125 kinetics among patients with advanced disease revealed a steeper decline in CA125 levels among LTS, suggesting reduced disease burden and improved treatment response. Comparisons involving LTS >10 years did not reach statistical significance, likely due to limited sample size (N = 23).

Contrary to prior clinical trial data suggesting a survival benefit of bevacizumab in high-risk ovarian cancer, our real-world analysis did not demonstrate a significant association between bevacizumab use and LTS in this subgroup.^27–29 Notably, large, randomized trials such as ICON-7 and GOG-0218 reported improvements in PFS and in some subgroups prolonged OS, particularly among patients with high-risk features. However, these trials were conducted under highly controlled settings, with protocol-driven patient selection and standardized treatment delivery.

Our study reflects real-world clinical practice, where bevacizumab was administered selectively to patients with advanced-stage or residual disease. The absence of a significant survival benefit observed in our cohort may reflect a combination of factors, including variability in the timing and duration of bevacizumab treatment, and unmeasured confounders such as comorbidities, performance status, or treatment adherence. Additionally, sample size limitations and wide confidence intervals raise the possibility of an underpowered analysis, given the relatively small number of high-risk patients treated with bevacizumab across LTS groups: those surviving <2 years vs >5 (n = 168), >7 (n = 117), and >10 years (n = 93). These findings suggest that, although commonly used in high-risk cases, bevacizumab was not independently predictive of LTS in this cohort. The results highlight the importance of evaluating therapeutic interventions outside of clinical trials and that patients’ BRCA status and cytoreductive outcomes as more consistent and robust predictors of long-term survival.

Our study has several limitations. First, its retrospective, single-center design may limit the generalizability of the findings. Second, access to HRD testing and PARP inhibitor therapy was limited during the study period. Poly (ADP-ribose) polymerase inhibitors were approved by the Israeli Ministry of Health in 2016, with reimbursement for second-line maintenance initiated in 2017 and first-line maintenance reimbursement only beginning in 2021. Consequently, the follow-up duration in the LTS dataset was insufficient to assess the long-term impact of first-line maintenance PARP inhibitors, precluding meaningful evaluation of their association with 10-year survival outcomes. Although some patients received PARP inhibitors earlier as part of clinical trials, extended survival data were not available to enable robust comparisons between first-line and recurrent PARP inhibitor administration. Additionally, we were unable to stratify outcomes based on germline BRCA PV vs HRD status due to limited sample size and incomplete genomic data in some cases. Most patients did not undergo HRD testing or receive PARP inhibitors, which may have impacted treatment outcomes and survival analyses. These factors should be considered when interpreting the results. These limitations highlight the need for long-term, prospective data to more accurately evaluate the effect of PARP inhibitors across treatment settings and biomarker-defined subgroups.

Moreover, in our evaluation of CA125 kinetics—specifically the slope of decline following chemotherapy—and its association with LTS, we observed a trend suggesting potential utility in identifying LTS. Although a trend toward association between the CA125 slope and LTS was observed, the analysis was constrained by the small number of patients with sufficient serial CA125 measurements. In many cases, CA125 values were not consistently available within the first 100 days of treatment, as required by established CA125 elimination rate constant K (KELIM) models.^30,31 Consequently, neither baseline CA125 levels nor the slope of decline were included in the final prognostic calculator, which was designed for use at or near the time of diagnosis. Future studies incorporating standardized and early longitudinal CA125 assessments are needed to further evaluate the prognostic value of dynamic CA125 changes.

The prognostic calculator was developed using variables that were statistically significant in multivariable logistic regression models, supporting its internal consistency and baseline validity within the study cohort. Nonetheless, external validation using multi-institutional datasets is needed, including assessments of calibration, area under the curve, and other performance metrics. As the model’s weights and intercepts were derived from a single-institution cohort, broader validation is necessary to confirm its generalizability and predictive performance across diverse patient populations.

Despite these limitations, the study’s strengths include a large patient cohort with long follow-up and comprehensive genetic testing for BRCA. This work proposes a novel practical tool for clinicians, offering valuable prognostic information. The prognostic calculators may support clinical decision-making and facilitate tailored treatments based on individual risk profiles. Future prospective studies are needed to validate these results and ensure their broad applicability.

Author contributions

Eliya Shachar (Conceptualization, Data curation, Investigation, Methodology, Writing—original draft, Writing—review & editing), Yael Raz (Investigation, Resources), Gilat Rotkop (Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Writing—review & editing), Bar Levy (Data curation, Investigation), Adi Diner (Data curation), Ido Laskov (Data curation, Resources), Nadav Michan (Resources), Dan Grisaru (Resources, Writing—review & editing), Ido Wolf (Conceptualization, Methodology, Writing—review & editing), and Tamar Safra (Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Supervision, Writing—review & editing)

Supplementary material

Supplementary material is available at The Oncologist online.

Funding

E.S. received T32 National Institutes of Health (NIH) funding for the Patient-Centered Outcomes Research Training in Urologic and Gynecologic Cancers (PCORT UroGynCan) grant number T32CA251072.

Conflicts of interest

I.W. received Honoraria (paid lectures): BMS, AZ, Merck-­Serono, Pfizer.

Data availability

The data underlying this article will be shared on reasonable request to the corresponding author.