2.1 Patients

This retrospective study reviewed the medical records of patients with gynecological cancer who received combination therapy with triweekly pembrolizumab (100 or 200 mg) and bevacizumab (7.5–15 mg/kg) between January 2018 and December 2021 at Shuang Ho Hospital, Taipei Medical University. Eligible patients comprised those with recurrent EOC who had undergone at least two or more prior lines of chemotherapy and had an Eastern Cooperative Oncology Group (ECOG) performance status of 0–2. Patients with incomplete follow-up data, including missing CA-125 measurements or imaging records, were excluded to minimize bias. The last follow-up date was 31 December 2021. All patients provided informed consent for study participation, and data were anonymized in accordance with institutional guidelines. The study was conducted under protocols approved by the Institutional Review Board of Shuang Ho Hospital, Taipei Medical University (Reference No. N202005105).

2.2 Outcome measurements

The following clinical parameters were extracted from medical records: age, ECOG performance status, disease stage at diagnosis, histological subtype, PD-L1 status, platinum sensitivity or resistance, prior lines of therapy, total treatment cycles, adverse effects and current patient status. The disease stage was categorized as early-stage (I–II) or late-stage (III–IV). PD-L1 status was assessed by a pathologist using immunohistochemistry (IHC), with positivity defined as membrane staining in at least 1% of viable tumor cells at any intensity. The SP263 assay (Ventana) was performed on the BenchMark XT platform, using the OptiView DAB IHC Detection Kit for visualization. The primary outcome was the ORR. Secondary outcomes included progression-free survival (PFS) and overall survival (OS). PFS was defined as the time from initiation of combination therapy to disease progression, determined by either CA-125 progression or immune-related Response Evaluation Criteria in Solid Tumors (irRECIST), or death from any cause in the absence of progression. Patients who remained alive without disease progression were censored at their last recorded assessment. OS was defined as the time from treatment initiation to death from any cause or last follow-up. For patients who were alive at the last follow-up, OS was censored at the most recent date they were known to be alive.

2.3 Response rates

The response rates were evaluated based on the Gynecological Cancer Intergroup (GCIG) criteria using serum CA-125 levels, categorized into four groups: response with normalization, response, non-response and progression [16]. A CA-125 response was defined as at least a 50% reduction from baseline levels in a pretreatment sample. The patients were classified as having response with normalization if their CA-125 levels both met the response criteria and decreased to within the reference range. CA-125 progression was defined as an increase in CA-125 levels to at least twice the nadir value on two separate occasions at least one week apart. Non-response was defined as any change in CA-125 levels that did not meet the criteria for response or progression. The ORR was calculated as the sum of the rates of response with normalization and CA-125 response. In addition to CA-125 assessments, imaging data from abdominal and pelvic computed tomography (CT) or magnetic resonance imaging (MRI) were collected when available for tumor follow-up. Imaging assessments were independently reviewed by a radiologist blinded to patient clinical data. Response rates were further evaluated using the irRECIST [17]. Patients with less than two available imaging data during the period of combination therapy cannot be evaluated using the irRECIST. For patients whose pretreatment CA-125 levels were within the normal range and did not meet GCIG criteria, response rates were determined using irRECIST.

2.4 Statistical analysis

Categorical variables were compared using the chi-squared test and are presented as absolute numbers and percentages. Continuous variables were analyzed using the Student’s t-test and are reported as means, standard deviations, and ranges (minimum to maximum). A two-tailed p-value < 0.05 was considered statistically significant. For PFS and OS, median survival estimates and corresponding 95% confidence intervals (CIs) were derived using Kaplan-Meier curves, with CIs calculated based on Greenwood’s formula. Given the small sample size, survival analyses should be interpreted with caution. Further studies with larger cohorts are necessary to validate these preliminary findings.

3.1 Patient characteristic

A total of 12 patients were included in this study, comprising six (50.0%) with high-grade serous carcinoma, including one case of peritoneal serous papillary carcinoma; three (25.0%) with clear cell carcinoma; one (8.3%) with mucinous carcinoma; one (8.3%) with endometrioid carcinoma; and one (8.3%) with mixed histology (95% dedifferentiated and 5% clear cell carcinoma) (Table 1, Supplementary Fig. 1). All patients had recurrent disease and had undergone at least two to three prior lines of therapy.

The mean number of previous chemotherapy regimens was 3.8, with seven patients (58.3%) having received more than three lines of chemotherapy. The mean patient age was 55.8 years (range, 39–69), and nine patients (75.0%) were diagnosed at an advanced stage (International Federation of Gynecology and Obstetrics (FIGO) stage III–IV). Regarding PD-L1 status, six patients (50.0%) had negative expression, and three (25.0%) were positive. PD-L1 expression could not be assessed in three patients due to the unavailability of paraffin-embedded tissue samples at the study institution. Among the PD-L1-positive cases, expression levels in tumor cells were low, with positivity observed in 1% (clear cell carcinoma patient), 1% (high grade serous carcinoma patient), and 5% (high grade serous carcinoma patient) of tumor cells, respectively.

The median baseline CA-125 level was 476.8 U/mL. The mean number of combination therapy cycles administered was 4.6 (range, 1–15). At the last follow-up, six patients had died, while six remained alive.

3.2 Efficacy analysis

We observed a reduction in CA-125 levels after the first cycle of combination therapy was observed in nine patients (75.0%). Based on the GCIG criteria, one patient (8.3%) achieved a response with normalization, three patients (25.0%) demonstrated a response, four patients (33.3%) showed no response, and three patients (25.0%) exhibited disease progression (Table 2). One patient was considered not assessable according to the GCIG criteria due to a pretreatment CA-125 level within the normal range (13.9 U/mL). This patient was instead evaluated using irRECIST and demonstrated a partial response. The ORR, calculated as the sum of response with normalization and response, including the irRECIST partial response case, was 41.7%.

Among patients with clear cell carcinoma, two of the three (66.7%) exhibited a response, including one who achieved response with normalization. This patient was initially diagnosed with FIGO stage II clear cell carcinoma in 2014, experienced multiple recurrences and had undergone at least three lines of adjuvant therapy before initiating pembrolizumab and bevacizumab therapy in 2019. Before treatment, her CA-125 level was 1236.6 U/mL, and after one cycle of combination therapy, her CA-125 levels declined by nearly 50%. Continued triweekly treatment resulted in a further decrease, reaching the normal range (<35 U/mL) after seven cycles. A CT scan confirmed tumor regression, and a complete response was achieved after nine cycles. She completed 12 cycles of combination therapy and remained under follow-up at the outpatient department [16].

In July 2021, a recurrent tumor was detected on CT scan (Fig. 1A), accompanied by an elevated CA-125 level (51.4 U/mL). The patient was reinitiated on pembrolizumab and bevacizumab therapy. One month after re-treatment, a CT scan showed complete remission (Fig. 1B), and CA-125 levels returned to the normal range. At the time of manuscript preparation, the patient remained disease-free.

Fig. 1.CT scan before and after dual combination therapy. This figure illustrates the radiological response of a patient with recurrent clear cell carcinoma who achieved a complete response and normalization of CA-125 following treatment with pembrolizumab and bevacizumab. (A) Recurrent pelvic tumors (blue arrow) detected in July 2021. (B) Significant regression of recurrent pelvic tumors observed one month after initiation of pembrolizumab and bevacizumab therapy.

3.3 Best tumor response

The best tumor response was determined based on the maximum percentage change in CA-125 levels from baseline (Fig. 2). One patient without CA-125 evaluation was excluded from this analysis. Among the six patients with high-grade serous carcinoma, three exhibited tumor progression, characterized by an increase in CA-125 levels from baseline. In contrast, all three patients with clear cell carcinoma demonstrated tumor regression. As shown in Fig. 2, two of the three OCCC patients who responded to combination therapy are clearly identifiable: one achieved response with normalization, while the other demonstrated a response. In contrast, all six patients with serous carcinoma, including one with peritoneal serous papillary carcinoma, exhibited either non-response or progression following treatment. The observed differences in response rates highlight the potential influence of histological subtype, with clear cell carcinoma demonstrating a higher response rate compared to serous carcinoma. The response data according to histological type and PD-L1 status can be found in Supplementary Table 1.

Fig. 2.Best tumor response based on CA-125 change from baseline. This figure illustrates the best tumor response, calculated as the maximum fold change in CA-125 levels relative to baseline. One patient without CA-125 evaluation was excluded. Among the six patients with high-grade serous carcinoma, three exhibited tumor progression, characterized by an increase in CA-125 levels from baseline. In contrast, all three patients with clear cell carcinoma demonstrated tumor regression. Abbreviations: CCC: clear cell carcinoma; EN: endometrioid carcinoma; MU: mucinous carcinoma; HG: high-grade serous carcinoma; PSPC: peritoneal serous papillary carcinoma; NR: normalized; CA-125: cancer antigen 125.

According to the irRECIST, two patients (16.7%) achieved an irRECIST partial response (irPR), four patients (33.3%) showed irRECIST stable disease (irSD), and three patients (25.0%) presented with irRECIST progressive disease (irPD) (Supplementary Table 1). The remaining three patients could not be assessed due to having fewer than two recorded imaging data during combination therapy treatment period. The responses of patients evaluated using the GCIG and irRECIST criteria are correlated in Supplementary Fig. 2, indicating compatibility between the two assessment methods.

3.4 Progression-free survival and overall survival

The median PFS for the study population was 2.8 months (mean PFS of 6.4 months, 95% CI, 0.7–12.2). Patients with OCCC exhibited a more favorable median PFS of 11.6 months (mean PFS of 15.1 months, 95% CI, −22.2 to 52.4) compared to median PFS of 2.6 months (mean PFS of 3.5 months, 95% CI, 0.4–6.7) in non-OCCC patients (p = 0.05, Fig. 3A). The median OS was 12.8 months (mean OS of 14.5 months, 95% CI, 7.7–21.4). A trend toward improved OS was observed in OCCC patients, with a median OS non-reachable (mean OS of 22.7 months, 95% CI, −2.6 to 47.9) compared to a median OS of 6.6 months (mean OS of 11.8 months, 95% CI, 4.1–19.5) in non-OCCC patients (p = 0.07, Fig. 3B). The small sample size of 12 patients limits the ability to draw definitive conclusions regarding survival outcomes. Larger prospective studies are necessary to validate these findings and further investigate the potential benefits of pembrolizumab and bevacizumab in recurrent ovarian cancer, particularly in clear cell carcinoma.

Fig. 3.Progression-free survival and overall survival. (A) The median PFS for the study population was 2.8 months (mean PFS of 6.4 months, 95% CI, 0.7–12.2). Patients with OCCC exhibited a more favorable median PFS of 11.6 months (mean PFS of 15.1 months, 95% CI, −22.2 to 52.4) compared to median PFS of 2.6 months (mean PFS of 3.5 months, 95% CI, 0.4–6.7) in non-OCCC patients (p = 0.05). (B) The median OS was 12.8 months (mean OS of 14.5 months, 95% CI, 7.7–21.4). A trend toward improved OS was observed in OCCC patients, with a median OS non-reachable (mean OS of 22.7 months, 95% CI, −2.6 to 47.9) compared to a median OS of 6.6 months (mean OS of 11.8 months, 95% CI, 4.1–19.5) in non-OCCC patients (p = 0.07). CCC: clear cell carcinoma.

3.5 Drug-related adverse events

Drug-related adverse events of grade 1–2 were observed in five patients (41.7%), while grade 3 adverse events occurred in four patients (33.3%) (Supplementary Table 2). No grade 4 adverse events were reported. The most common grade 1–2 drug-related adverse events included nausea (2 patients, 16.7%), diarrhea (2 patients, 16.7%), rash (2 patients, 16.7%), pruritus (2 patients, 16.7%), and dyspnea (3 patients, 25.0%). The most frequently observed grade 3 drug-related adverse event was lymphopenia (3 patients, 25.0%). No drug-related adverse event resulted in treatment interruption or discontinuation.

4.1 Summary of main results

This retrospective study evaluated 12 patients with heavily treated recurrent EOC who received combination therapy with pembrolizumab and bevacizumab, revealing an ORR of 41.7%, with 8.3% achieving response with normalization and 33.3% demonstrating a response. Patients with OCCC exhibited the most favorable response, with an ORR of 66.7%, including one patient who achieved response with normalization. A reduction in CA-125 levels was observed after the first treatment cycle in 75% of patients. The adverse events associated with this regimen were predominantly mild, including nausea, diarrhea and rash, all occurring at lower rates than those reported in previous studies. While a prior phase II trial involving a similar combination regimen reported 42.5% of patients requiring dose interruptions and 10% experiencing treatment discontinuation due to adverse events [12], no patients in the present study required treatment interruption or discontinuation, suggesting that the combination of pembrolizumab and bevacizumab was well tolerated, with manageable side effects that did not significantly impact treatment adherence. Together, these findings indicate that the combination of an angiogenesis inhibitor and immunotherapy may offer a promising therapeutic approach for recurrent EOC, particularly OCCC. However, given the small sample size, the results should be interpreted with caution, and further validation in larger, prospective studies is necessary to confirm the efficacy and safety of this treatment strategy.

4.2 Results in the context of published literature

Bevacizumab (trade name: Avastin) is an angiogenesis inhibitor that blocks vascular endothelial growth factor A (VEGF-A) and has been incorporated into chemotherapy regimens as a maintenance strategy or in combination therapy across various ovarian cancer settings. Several phase III trials (i.e., GOG-0218, OCEANS, AURELIA, ICON7) have demonstrated its efficacy in front-line, recurrent platinum-sensitive and platinum-resistant ovarian cancer [18, 19]. However, single agent bevacizumab in heavily pretreated recurrent EOC remains limited, which demonstrated modest response rates of 21% and 15% in previous studies [20, 21]. Given the limited efficacy of bevacizumab monotherapy, combining it with an ICI is a rational therapeutic strategy. Bevacizumab may modulate the tumor microenvironment by normalizing vasculature and enhancing immune cell infiltration, potentially improving the activity of ICIs. The hypothesized synergy between anti-angiogenic therapy and immunotherapy forms the basis for our use of this combination to increase treatment responsiveness in recurrent EOC. Pembrolizumab (Keytruda), a humanized monoclonal antibody targeting PD-1 on lymphocytes, has been approved by the US Food and Drug Administration (FDA) for treating multiple malignancies, including melanoma, non-small cell lung cancer, head and neck squamous cell carcinoma, Hodgkin lymphoma, urothelial carcinoma, gastric cancer and tumors with mismatch repair deficiency or microsatellite instability. However, pembrolizumab has not been approved for ovarian cancer [22]. The phase II KEYNOTE-100 trial (ClinicalTrials.gov identifier: NCT02674061) enrolled 376 patients with recurrent ovarian cancer to assess pembrolizumab monotherapy [10], and reported that the ORR was limited to 8.0%, with 1.9% achieving complete response and 6.1% achieving partial response. Notably, in a subgroup analysis of 21 Japanese patients with OCCC, the ORR was higher at 19.0% (95% CI: 5.4–41.9), suggesting a potential benefit in this histological subtype [23]. Although PD-L1 expression is considered a predictive biomarker for pembrolizumab response in the KEYNOTE-100 study, a heavily pre-treated patient with PD-L1-negative OCCC achieved complete remission following combination therapy with pembrolizumab and bevacizumab [24]. Several studies have identified biomarkers and immune checkpoint-genes that are abnormally expressed in OCCC, including cytotoxic T lymphocyte-associated protein 4 (CTLA4), PD-1, PD-L1, Lymphocyte-activation gene 3 (LAG 3) and T-cell immunoglobulin and mucin-domain containing-3 (Tim3). These findings suggest potential targets for ICIs therapy, indicating that patients with OCCC may derive benefits from such treatment [25]. Molecular alterations such as AT-rich interactive domain-containing protein 1A (ARID1A) mutations, overexpression of VEGF, annexin A4 and mammalian target of rapamycin (mTOR) pathway activation have been frequently reported in OCCC [26]. Among these molecular targets, an essential component of the SWItch/Sucrose NonFermentable (SWI/SNF) chromatin remodeling complex, has been found to play a role in mismatch repair by recruiting MutS Homolog 2 (MSH2) to chromatin during DNA replication [27], with ARID1A inactivation reported to compromise mismatch repair, leading to increased mutagenesis and neoantigen formation [28].

A clinical cohort study demonstrated that patients with ARID1A-altered tumors (n = 46) had significantly longer PFS after immune checkpoint blockade therapy compared to patients with wild-type ARID1A (n = 329) (p = 0.006) [29]. Multivariate analysis further confirmed that ARID1A mutations independently predicted improved PFS after immune checkpoint blockade (95% CI: 0.39–0.94, p = 0.02), irrespective of microsatellite instability (MSI) status or overall mutational burden [25]. In our study, the tumor from patient CCC3 harbored two ARID1A frameshift mutations (p.M1154fs7* and p.Y788*), which may have contributed to the observed therapeutic response. These findings suggest that a precision medicine approach, leveraging ARID1A loss-of-function mutations, may enhance the efficacy of ICIs, particularly in EOC and OCCC. Further research is warranted to explore the role of ARID1A alterations in optimizing immunotherapy strategies for ovarian cancer.

A phase II open-label clinical trial evaluated the combination of pembrolizumab, bevacizumab and oral cyclophosphamide in 40 patients with platinum-sensitive or platinum-resistant recurrent epithelial ovarian, fallopian tube, or primary peritoneal cancer [12], reporting a 6-month PFS rate of 70%, with 15 patients achieving a partial response and 13 patients maintaining stable disease. A subsequent retrospective study assessed the efficacy of the same triple combination regimen in patients with recurrent, platinum-resistant, heavily pre-treated ovarian cancer. The study reported an ORR of 13% and a disease control rate (DCR) of 33%, with a median PFS of 3.5 months [13]. Additionally, a phase I open-label trial investigated pembrolizumab and bevacizumab, with or without pegylated liposomal doxorubicin, in patients with platinum-resistant EOC who had received multiple prior lines of therapy, and reported that the combination of pembrolizumab and bevacizumab alone resulted in a promising ORR of 26.3% and a DCR of 78.9%, with a median PFS of 4.7 months [15]. The treatment was well tolerated, with no dose-limiting toxicities, suggesting that anti-PD-1 and anti-angiogenic agent combinations could serve as a potential therapeutic strategy for platinum-resistant EOC patients with limited treatment options.

4.3 Strengths and weaknesses

One of the strengths of this study is that despite the omission of oral cyclophosphamide due to concerns regarding its chemotherapy-related adverse effects, a comparable ORR was observed relative to Dr. Zsiros’s findings [12]. The favorable tolerability profile of this regimen may be attributed to the use of bevacizumab, which has a low toxicity profile, and the absence of cyclophosphamide, potentially making this combination more suitable for patients with heavily pre-treated recurrent EOC.

However, this study has several limitations. First, as a retrospective study, it is subject to potential selection biases and lacks a control group, which limits the ability to draw definitive conclusions. Second, the high medication costs and variability in insurance coverage restricted the sample size, reducing the statistical power of the findings. Consequently, while a trend toward improved survival was observed, the results did not reach statistical significance. Third, somatic genetic testing was not routinely performed in all EOC patients, preventing the establishment of an association between molecular alterations such as ARID1A mutations, and treatment response. To further assess the efficacy of the pembrolizumab and bevacizumab combination in recurrent OCCC and other EOC subtypes, larger prospective studies with comprehensive genomic profiling are warranted.

4.4 Implications for practice and future research

This study highlights two key aspects: the synergistic effects of combination therapy and the abscopal effect observed in select cases. Patient CCC3 initially received 12 cycles of combination therapy, achieving normalized CA-125 levels. However, 14 months later, tumor recurrence was detected along with an elevated CA-125 level. Upon re-treatment with pembrolizumab and bevacizumab, a complete remission was observed within one month, as confirmed by CT imaging (Fig. 1), and CA-125 levels returned to the normal range, which suggests that recurrent tumors may still respond to re-treatment with this combination, potentially offering a viable therapeutic strategy for recurrent disease. High-dose chemotherapy frequently results in severe complications and the development of chemoresistance, with more than 70% of patients experiencing relapse and eventual resistance to conventional chemotherapy [30]. By targeting two distinct pathways, combination therapies may enhance treatment sensitivity compared to monotherapy [31]. Bevacizumab, through its anti-angiogenic properties, has been shown to improve lymphocyte trafficking and restore dendritic cell function, thereby depleting circulating regulatory T cells and enhancing the cytotoxic activity of T cells and natural killer cells [32]. These mechanisms may contribute to the increased efficacy of pembrolizumab in combination with bevacizumab. Furthermore, polyadenosine diphosphate ribose polymerase inhibitors (PARPi) have shown significant advancements in ovarian cancer treatment and may modulate the immune microenvironment by upregulating PD-L1 expression, activating immune pathways, and increasing genomic instability [33]. These effects suggest that the combination of PARPi with ICIs warrants further investigation as a potential therapeutic strategy.

A second notable case in this study involved a patient with mixed epithelial ovarian carcinoma (95% dedifferentiated and 5% clear cell, FIGO stage IIIB), who had disease recurrence following three lines of adjuvant chemotherapy. Despite receiving pembrolizumab and bevacizumab, interval progression of recurrent tumors was observed on follow-up imaging after three months. As a result, radiotherapy was introduced to target hepatic metastases. Following treatment, tumor regression was observed after four months, consistent with the abscopal effect, which describes a systemic tumor response triggered by localized radiotherapy (Supplementary Fig. 3). The abscopal effect suggests that low-dose radiation may reprogram the tumor microenvironment, reversing immune desertification and potentially overcoming resistance to immunotherapy [34]. Radiotherapy has been increasingly recognized as a trigger for systemic antitumor immune responses, and its integration into combination regimens could enhance the efficacy of ICIs and anti-angiogenic agents, particularly in metastatic and treatment-resistant cancers [35].

The findings from this study suggest potential extensions of pembrolizumab and bevacizumab therapy in recurrent ovarian cancer, particularly through re-treatment strategies, synergistic combination approaches, and integration with radiotherapy. Further investigations are needed to evaluate the long-term efficacy, optimal sequencing and molecular predictors of response to combination immunotherapy in ovarian cancer.