Pembrolizumab plus weekly paclitaxel in platinum-resistant recurrent ovarian cancer (ENGOT-ov65/KEYNOTE-B96): a multicentre, randomised, double-blind, phase 3 study.

# Pembrolizumab plus weekly paclitaxel in platinum-resistant recurrent ovarian cancer (ENGOT-ov65/KEYNOTE-B96): a multicentre, randomised, double-blind, phase 3 study *The Lancet 2026* Articles Pembrolizumab plus weekly paclitaxel in platinum-resistant recurrent ovarian cancer (ENGOT-ov65/KEYNOTE-B96): a multicentre, randomised, double-blind, phase 3 study Nicoletta Colombo*, Emese Zsiros*, Gabriella Parma, Eliana Rulli, Alexandra Sebastianelli, Mariusz Bidzinski, Carlos Gallardo, Emad Matanes, Kosei Hasegawa, Fatih Kose, Manuel Magallanes-Maciel, Rebecca A Herbertson, Sumitra Ananda, Judith R Kroep, Andreia Cristina de Melo, Philip R Debruyne, Jae-Weon Kim, Jalid Sehouli, Marc-Edy Pierre, Sakari Hietanen, Claudio Zamagni, Xin Lu, Bradley J Monk, Robert L Coleman, Xuan Peng, Karin Yamada, Agata M Bogusz, Thibault De La Motte Rouge, Xiaohua Wu, on behalf of the ENGOT-ov65/KEYNOTE-B96 investigators† Summary Background Epithelial ovarian cancer frequently recurs and becomes resistant to platinum chemotherapy. We Lancet 2026; 407: 1525–37 investigated whether adding pembrolizumab to weekly paclitaxel, with or without bevacizumab, improves progression- Published Online free survival and overall survival compared with weekly paclitaxel, with or without bevacizumab, in participants with April 10, 2026 platinum-resistant recurrent ovarian cancer who had received one to two previous systemic regimens. https://doi.org/10.1016/ S0140-6736(26)00602-1 Methods ENGOT-ov65/KEYNOTE-B96 is a randomised, double-blind, phase 3 study conducted at 187 gynaecologic *Contributed equally oncology centres in 25 countries in the Americas, Asia, Europe, and Oceania. Adults (≥18 years) with histologically †ENGOT-ov65/KEYNOTE-B96 investigators are listed in the confirmed epithelial ovarian, fallopian tube, or primary peritoneal carcinoma, who received one to two previous appendix systemic therapies including at least one platinum regimen and who progressed 6 months or less after the last Gynecologic Oncology Program platinum regimen, were eligible. Participants were randomly assigned 1:1 to intravenous pembrolizumab 400 mg (N Colombo MD PhD) and every 6 weeks for up to 18 cycles plus open-label intravenous paclitaxel 80 mg/m² on days 1, 8, and 15 of each 21-day Division of Gynecologic cycle or intravenous placebo (saline solution) every 6 weeks for up to 18 cycles plus open-label intravenous paclitaxel Oncology (Gabriella Parma MD), European Institute of 80 mg/m² on days 1, 8, and 15 of each 21-day cycle; intravenous bevacizumab 10 mg/kg every 2 weeks was permitted Oncology, IRCCS, Milan, Italy; per investigator. Randomisation was stratified by planned bevacizumab use, region, and PD-L1 combined positive Department of Gynecologic score (CPS). The primary endpoint was investigator-assessed progression-free survival per RECIST version 1.1; the Oncology, Roswell Park key secondary endpoint was overall survival. Results from two interim analyses and the final analysis are included in Comprehensive Cancer Center, Buffalo, NY, USA this Article. This study is registered with ClinicalTrials.gov, NCT05116189, and is now completed. (Prof E Zsiros MD PhD); Laboratory of Methodology for Findings Between Dec 13, 2021, and July 3, 2023, 643 female participants were randomly assigned; 322 to Clinical Research, Clinical pembrolizumab plus paclitaxel and 321 to placebo plus paclitaxel. At the first interim analysis, pembrolizumab plus Oncology Department, Istituto di Ricerche Farmacologiche paclitaxel significantly improved progression-free survival versus placebo plus paclitaxel in both the PD-L1 CPS 1 or Mario Negri IRCCS, Milan, Italy higher (median 8·3 months vs 7·2 months; hazard ratio [HR] 0·72, 95% CI 0·58–0·89; p=0·0014 [α=0·012]) and (E Rulli PhD); CHU de Québec- overall populations (median 8·3 months vs 6·4 months; HR 0·70, 95% CI 0·58–0·84; p<0·0001 [α=0·0023]), meeting Université Laval, QC, Canada (A Sebastianelli MD FRCS); the prespecified criteria for confirmatory efficacy. At the second interim analysis, overall survival was significantly Narodowy Instytut Onkologii improved in the PD-L1 CPS 1 or higher population (median 18·2 months vs 14·0 months; HR 0·76, 95% CI 0·61–0·94; im Marii Skłodowskiej-Curie, p=0·0053 [α=0·0083]). At the final analysis, overall survival was significantly improved in the overall population Warsaw, Poland (median 17·7 months vs 14·0 months; HR 0·82, 95% CI 0·69–0·97; p=0·011 [α=0·024]). Grade 3 or worse treatment- (Prof M Bidzinski MD PhD); Bradford Hill Clinical Research related adverse events occurred in 217 (68%) of 320 participants in the pembrolizumab plus paclitaxel group versus Center and Finis Terrae 176 (55%) of 318 participants in the placebo plus paclitaxel group. The most common treatment-related adverse University, Santiago, Chile events (any grade) included anaemia, peripheral neuropathy, alopecia, fatigue, and nausea. Treatment-related adverse (C Gallardo MD); Rambam events resulted in death in four participants (1%) in the pembrolizumab plus paclitaxel group (colitis, interstitial lung Health Care Campus, and Ruth and Bruce Rappaport Faculty of disease, acute myeloid leukaemia, and intestinal perforation) and in five participants (2%) in the placebo plus Medicine, Technion, Haifa, paclitaxel group (cardiac failure, intestinal perforation [in two participants], and large-intestine perforation [in Israel (E Matanes MD); Saitama two participants]). Medical University International Medical Center, Hidaka, Saitama, Japan Interpretation Pembrolizumab plus weekly paclitaxel, with or without bevacizumab, significantly improved (Prof K Hasegawa MD PhD); progression-free survival and overall survival in participants with platinum-resistant recurrent ovarian cancer who Başkent University, Ankara, had received one to two previous systemic regimens, supporting this regimen as a new treatment option for this Turkey (Prof F Kose MD); Centro population. Oncologico Internacional, Mexico City, Mexico (M Magallanes-Maciel MD); Funding Funded by Merck Sharp & Dohme LLC, a subsidiary of Merck & Co, Inc, Rahway, NJ, USA. University Hospitals Sussex NHS Foundation Trust, Copyright © 2026 Elsevier Ltd. All rights reserved, including those for text and data mining, AI training, and similar Worthing, UK technologies. Articles (R A Herbertson DM); Epworth Healthcare and Peter Research in context MacCallum Cancer Centre, Evidence before this study largely negative results in this setting. To our knowledge, no Melbourne, VIC, Australia (S Ananda MBBS FRACP); Leiden We searched PubMed, major conference proceedings, and previous randomised phase 3 study had demonstrated an University Medical Center, ClinicalTrials.gov up to Jan 31, 2026, for phase 2–3 studies overall survival benefit with an immune checkpoint inhibitor– Leiden, Netherlands and Dutch evaluating immune checkpoint inhibitors for platinum- based regimen in ovarian cancer and the potential benefit of Gynaecology Oncology Group (Prof J R Kroep MD PhD); resistant recurrent ovarian cancer using the keywords: “ovarian combining pembrolizumab and weekly paclitaxel had not been Division of Clinical Research cancer”, “platinum-resistant”, “recurrent ovarian cancer”, validated in a definitive trial. and Technological “immune checkpoint inhibitor”, “pembrolizumab”, “paclitaxel”, Development, Brazilian Added value of this study “metronomic paclitaxel”, “phase 2”, and “phase 3”. In the National Cancer Institute, Rio The randomised, placebo-controlled phase 3 ENGOT-ov65/ de Janeiro, Brazil KEYNOTE-028 and KEYNOTE-100 studies, pembrolizumab KEYNOTE-B96 study showed that adding pembrolizumab to (A C de Melo MD PhD); Kortrijk monotherapy showed modest activity in recurrent ovarian weekly paclitaxel, with or without bevacizumab, significantly Cancer Centre, az groeninge, cancer, with objective response rates of 8–12% and greater Kortrijk, and Belgium and prolongs progression-free survival and overall survival in benefit in PD-L1-expressing tumours. Several investigator- Luxembourg Gynaecological platinum-resistant recurrent ovarian cancer in both the PD-L1 Oncology Group, Leuven, initiated studies explored combination immunotherapy combined positive score 1 or higher and the overall Belgium strategies in this setting. A phase 2 study (NCT02440425) of populations, representing the first demonstration of an overall (Prof P R Debruyne MD PhD); weekly paclitaxel plus pembrolizumab suggested encouraging Medical Technology Research survival benefit with an immune checkpoint inhibitor regimen efficacy compared with historical weekly paclitaxel. In another Centre, School of Allied Health in ovarian cancer. The magnitude and consistency of benefit and Social Care, Anglia Ruskin phase 2 study (NCT02853318), pembrolizumab combined with across clinically relevant subgroups—including older patients, University, Chelmsford, UK bevacizumab and oral metronomic cyclophosphamide (Prof P R Debruyne); School of produced an objective response rate of 43% and a median those previously treated with poly(ADP-ribose) polymerase Nursing and Midwifery, inhibitors, and those with short platinum-free intervals— progression-free survival of 7·6 months in the platinum- University of Plymouth, underscore the robustness of these findings. Safety outcomes Plymouth, UK resistant cohort, supporting the rationale for combining were consistent with the known profiles of pembrolizumab and (Prof P R Debruyne); Seoul immunotherapy with immune-modulating or metronomic National University, Seoul, chemotherapy. weekly paclitaxel, with no new safety signals and manageable South Korea immune-mediated adverse events. (Prof J-W Kim MD PhD); Asia- Two randomised phase 3 trials, evaluating atezolizumab Pacific Gynecologic Oncology (AGO-OVAR 2.29/ENGOT-ov34, NCT03353831) or avelumab Implications of all the available evidence Trials Group, Seoul, (JAVELIN Ovarian 200, NCT02580058) added to single-agent Taken together with previous negative phase 3 trials of immune South Korea (Prof J-W Kim); Charite Universitaetsmedizin, chemotherapy (pegylated liposomal doxorubicin or weekly checkpoint inhibitors combined with non-metronomic Berlin, Germany and North- paclitaxel), did not show statistically significant improvements chemotherapy in platinum-resistant recurrent ovarian cancer, Eastern German Society of in progression-free survival or overall survival in unselected the findings from ENGOT-ov65/KEYNOTE-B96 indicate that Gynecological Oncology platinum-resistant recurrent ovarian cancer populations. the choice and schedule of chemotherapy are critical (Prof J Sehouli MD); Gynecology Oncology Unit, Luis Carlos However, exploratory analyses suggested higher activity when determinants of immunotherapy effectiveness. Weekly Samiento Angulo Cancer immune checkpoint inhibitors were combined with weekly (metronomic) paclitaxel appears to potentiate checkpoint Treatment and Research paclitaxel compared with other chemotherapy backbones. This blockade, consistent with translational data suggesting Center–CTIC, Bogotá, Colombia observation is consistent with preclinical evidence that enhanced antigen presentation, reduced immunosuppressive (M-E Pierre MD); Turku University Hospital and NSGO- metronomic or low-dose taxanes can enhance antigen cell populations, and a more immunogenic tumour CTU, Turku, Finland presentation, reduce immunosuppressive cell populations, and microenvironment with metronomic dosing. These results (Prof S Hietanen MD); IRCCS enhance the activity of immune checkpoint blockade. establish pembrolizumab plus weekly paclitaxel, with or Azienda Ospedaliero– without bevacizumab, as an effective treatment option for Universitaria di Bologna, Although between-trial differences preclude direct platinum-resistant recurrent ovarian cancer, provide the first Bologna, Italy (C Zamagni MD); comparisons, several recent studies in platinum-resistant Gynecologic Oncology ovarian cancer, MIRASOL (mirvetuximab soravtansine), SCORES randomised evidence of an overall survival benefit with Department, Obstetrics and immunotherapy in ovarian cancer, and support this regimen as (suvemcitug plus chemotherapy), and ROSELLA (relacorilant Gynecology Hospital, Fudan a new standard of care. University, Shanghai, China plus nab-paclitaxel), have shown significant improvements in (Prof X Lu MD PhD); Florida progression-free survival and overall survival after decades of Cancer Specialists and Research Institute, West Palm Beach, FL, USA (Prof B J Monk MD); Gynecologic Oncology, Texas Introduction combined with bevacizumab, and antibody–drug Oncology, US Oncology Epithelial ovarian cancer is the third most common and conjugates.4 In the phase 3 AURELIA study, the addition Network Shenandoah, TX, USA (Robert L Coleman MD); Merck & the most lethal gynaecologic cancer globally.1 The of bevacizumab to non-platinum chemotherapy Co, Rahway, NJ, USA majority of patients with advanced ovarian cancer improved median progression-free survival from (X Peng PhD, K Yamada MD MSc, experience disease recurrence and eventual resistance to 3·4 months to 6·7 months (hazard ratio [HR] 0·48 A M Bogusz MD PhD); Centre platinum-based chemotherapy.2,3 Patients with platinum- [95% CI 0·38–0·60]; p<0·001), although overall survival Eugene Marquis, Rennes, and GINECO, Paris, France resistant recurrent ovarian cancer have a poor prognosis was not significantly improved.5 In the prespecified (T De La Motte Rouge MD PhD); despite treatment with current standard therapies, weekly paclitaxel subgroup, bevacizumab was associated Fudan University Shanghai including non-platinum-based chemotherapy, often with a marked improvement in median progression-free 1526 Articles survival (3·9 months to 10·4 months; HR 0·46) and had received one to two previous systemic regimens, and Cancer Center, Shanghai, China increased objective response rates from 30·2% to 53·3%, to characterise the safety profile of this regimen. (Prof X Wu MD PhD) establishing weekly paclitaxel with bevacizumab as the Correspondence to: most active regimen in this setting.5 More recently, the Methods Dr Nicoletta Colombo, European Institute of Oncology IRCCS, phase 3 SCORES trial showed that adding the VEGF Study design and participants Milan 20141, Italy antibody suvemcitug to chemotherapy in platinum- ENGOT-ov65/KEYNOTE-B96 is a randomised, double- nicoletta.colombo@ieo.it resistant ovarian cancer significantly improved blind, placebo-controlled, phase 3 study conducted at See Online for appendix progression-free survival and overall survival compared 187 sites (gynaecologic oncology centres) in 25 countries with chemotherapy alone, although the magnitude of (Australia, Belgium, Brazil, Canada, Chile, China, overall survival benefit was modest and interpretation is Colombia, Denmark, Finland, France, Germany, Ireland, constrained by the specific trial population.6 Nonetheless, Israel, Italy, Japan, Mexico, South Korea, the Netherlands, effective treatments that improve long-term outcome New Zealand, Norway, Poland, Russia, Turkey, the UK, data in this population remain limited, underscoring the and the USA). Eligible participants were aged 18 years or importance of novel therapeutic approaches to improve older, had histologically confirmed epithelial ovarian, patient outcomes.7 fallopian tube, or primary peritoneal carcinoma (high- Immune checkpoint inhibition has shown modest grade serous or predominantly serous, low-grade serous, activity as monotherapy in recurrent ovarian cancer, with any-grade endometrioid, malignant mixed Müllerian objective response rates of approximately 8–12%; tumours [carcinosarcoma], or clear cell carcinoma; for responses might be higher in some biomarker-defined the purposes of this study and consistent with subgroups.8,9 In randomised phase 3 trials evaluating the contemporary clinical classification, epithelial ovarian, addition of immune checkpoint inhibitors to fallopian tube, and primary peritoneal carcinomas are chemotherapy in platinum-resistant recurrent ovarian collectively referred to as ovarian cancer), and had cancer, no significant improvements in progression-free received one to two previous lines of systemic therapy for survival or overall survival were observed in unselected ovarian cancer, including at least one previous platinum- populations.10,11 Exploratory analyses from one such study based therapy and four to nine cycles of first-line, of immune checkpoint inhibition plus bevacizumab and platinum-based therapy. Participants who had chemotherapy suggested that treatment efficacy can vary radiographic evidence of disease progression of 6 months by chemotherapy backbone, with greater activity observed or less after the last platinum regimen and those with with weekly paclitaxel than with pegylated liposomal primary platinum-refractory disease (radiographic doxorubicin.10 evidence of disease progression while receiving or within Preclinical and clinical data suggest that chemotherapy, 28 days of the last dose of first-line platinum-based such as paclitaxel, might enhance antitumour immune chemotherapy) were excluded from the study. responses by increasing tumour immunogenicity and Participants were also required to have an Eastern reducing immunosuppressive factors within the tumour Cooperative Oncology Group performance status score microenvironment, particularly when administered at a of 0 or 1 and to provide an archival or newly obtained lower dose on a weekly schedule.12–14 Consistent with this tumour tissue sample for assessment of PD-L1 status. rationale, previous single-arm phase 2 studies combining Full eligibility criteria are reported in the study protocol pembrolizumab with metronomic chemotherapy showed (appendix). During screening and when permitted by encouraging efficacy in platinum-resistant recurrent law, participants self-reported their sex as female, male, ovarian cancer. In one study, pembrolizumab plus weekly undifferentiated, or unknown, their race as one or more paclitaxel resulted in an objective response rate of 51·4% of American Indian or Alaska Native, Asian, Black or (range 34·4–68·1) and median progression-free survival African American, Native Hawaiian or Other of 7·2 months (95% CI 4·5–11·0).15 In another study, Pacific Islander, or White, and their ethnicity as Hispanic pembrolizumab combined with oral metronomic or Latino, Not Hispanic or Latino, or unknown. cyclophosphamide and bevacizumab yielded an objective This trial was conducted in accordance with the response rate of 43·3% (90% CI 29·6–58·2) and median standards of Good Clinical Practice and the Declaration progression-free survival of 7·6 months (90% CI of Helsinki. The trial protocol and all protocol 5·7–10·3) in the platinum-resistant cohort,16 further amendments were approved by the appropriate ethics supporting the inclusion of bevacizumab with committee at each participating institution. The ethics chemotherapy and pembrolizumab. committee for the lead principal investigator’s site that Based on these observations, we conducted the phase 3 approved the trial protocol was the Comitato Etico degli ENGOT-ov65/KEYNOTE-B96 study to determine IRCCS Istituto Europeo di Oncologia e Centro whether adding pembrolizumab to weekly paclitaxel, Cardiologico Monzino (approval number: R1542/21- with or without bevacizumab, improves progression-free IEO1631). Clinically important protocol deviations were survival and overall survival compared with weekly defined as those with the potential to compromise critical paclitaxel, with or without bevacizumab, in participants analyses of primary efficacy or safety endpoints, or to with platinum-resistant recurrent ovarian cancer who affect participant safety. Protocol deviations were Articles overseen by the sponsor, with centralised review Solid Tumours (RECIST) version 1.1 after completing the procedures to ensure consistent classification across first treatment course or if they had a confirmed complete treatment groups and study sites. Clinically important response. Participants could discontinue one study deviations from protocol-directed requirements are listed medication (pembrolizumab or placebo, paclitaxel or in the appendix (p 15). An external, independent data docetaxel, or bevacizumab), based on adverse events or monitoring committee oversaw the trial, assessed safety investigator clinical judgement while continuing the other approximately every 6 months, and assessed efficacy at medications in the regimen when appropriate; follow-up prespecified interim analyses. All participants provided continued per protocol. Crossover between treatment written informed consent before enrolment. groups was not permitted. This study is registered with ClinicalTrials.gov, Tumour imaging was performed at screening, every NCT05116189, and is now completed. 9 weeks through to week 54, and every 12 weeks thereafter until investigator-assessed radiographic disease progression Randomisation and masking per RECIST version 1.1, start of new anticancer treatment, Participants were randomly allocated in a 1:1 ratio to consent withdrawal, pregnancy, or death. Adverse events receive either pembrolizumab plus weekly paclitaxel (the were assessed from randomisation through to 30 days after pembrolizumab plus paclitaxel group) or placebo plus cessation of study treatment (90 days for serious adverse weekly paclitaxel (the placebo plus paclitaxel group). events) and graded according to National Cancer Institute Randomisation was performed centrally using an Common Terminology Criteria for Adverse Events interactive response system and was stratified according version 5.0. Patient-reported outcomes were assessed every to planned use of bevacizumab determined by the 6 weeks until treatment discontinuation. investigator before randomisation (yes vs no), region (USA vs EU vs rest of the world), and tumour PD-L1 Outcomes status (combined positive score [CPS] <1 vs CPS 1 to <10 The primary endpoint was investigator-assessed vs CPS ≥10). PD-L1 CPS thresholds of 1 and 10 were progression-free survival per RECIST version 1.1, defined prespecified based on previous pembrolizumab studies. as the time from randomisation to first documented The randomisation block size was four. disease progression or death due to any cause. The key As a double-blind study for pembrolizumab, the secondary endpoint was overall survival. Other participants, investigators, other study site staff (except prespecified secondary endpoints included progression- for the unmasked pharmacist), and sponsor personnel or free survival assessed by blinded independent central delegates were masked to pembrolizumab versus saline review and patient-reported outcomes using the European placebo administration. The unmasked pharmacist Organisation for Research and Treatment of Cancer provided study site staff with ready-to-use identically (EORTC) Quality of Life Questionnaire-Core 30 packaged pembrolizumab or saline infusion solutions (QLQ-C30) and the abdominal or gastrointestinal for administration at scheduled infusion visits. symptom scale of the EORTC Ovarian Cancer-Specific Quality of Life Questionnaire (QLQ-OV28). Procedures Adverse events and laboratory abnormalities were Participants received pembrolizumab 400 mg or placebo graded according to the National Cancer Institute intravenously every 6 weeks for up to 18 cycles plus open- Common Terminology Criteria for Adverse Events, label paclitaxel 80 mg/m² intravenously on days 1, 8, and 15 version 5.0. Immune-mediated adverse events were of each 3-week cycle. Participants could also receive open- programmatically determined from a predefined list of label bevacizumab 10 mg/kg intravenously every 2 weeks Medical Dictionary for Regulatory Activities terms. based on the investigator’s clinical judgement of baseline The original protocol, the final protocol, and a summary safety criteria, laboratory and physiologic parameters, of all endpoint changes are available in the appendix. concomitant medications, and other relevant considerations; bevacizumab was reimbursed by the study Statistical analysis sponsor to ensure continued access during the study. Efficacy was assessed in all randomly assigned Paclitaxel and bevacizumab were continued until disease participants. Progression-free survival and overall progression, prohibitive toxicity, or the maximum duration survival were estimated using the non-parametric allowed on the approved label or per local practice. Kaplan–Meier method. For primary progression-free Participants with a severe hypersensitivity reaction to survival, participants were censored at the last disease paclitaxel or with an adverse event requiring discontinuation assessment before either (1) two or more missed of paclitaxel could receive docetaxel 75 mg/m² intravenously assessments preceding an event, (2) initiation of new every 3 weeks instead of paclitaxel after consultation with anticancer treatment before documented progression, or the sponsor. Participants in the pembrolizumab group (3) their final assessment if no event or new treatment were eligible to receive an additional nine cycles of occurs, with the earliest applicable criterion used when pembrolizumab 400 mg every 6 weeks if they experienced multiple apply. For overall survival, participants without disease progression per Response Evaluation Criteria in documented death at the time of analysis were censored 1528 Articles at the date they were last known to be alive. Between- The second interim analysis was planned to occur group differences were assessed using the stratified approximately 20 months after the last participant was log-rank test, with the magnitude of the difference assigned, after approximately 389 progression-free assessed using the stratified Cox proportional hazards survival events had occurred in the PD-L1 CPS 1 or higher model. The randomisation stratification factors were population and approximately 541 progression-free applied to all stratified analyses. All statistical analyses survival events had occurred in the overall population. were performed with SAS (version 9.4). The consistency The final analysis was planned to occur after approximately of the treatment effect in prespecified subgroups with 369 deaths had occurred in the PD-L1 CPS 1 or higher the overall population was assessed descriptively using population and approximately 512 deaths had occurred in HRs and associated 95% CIs calculated using the non- the overall population. Formal hypothesis testing for stratified Cox proportional hazards model. progression-free survival was prespecified at the first and The study was designed to enrol approximately second interim analyses only, with the second interim 643 participants. Baseline characteristics were analysis constituting the final opportunity for α-controlled summarised descriptively without formal between-group testing of progression-free survival. If the progression-free testing. The trial design included two interim analyses survival hypothesis was rejected at the first interim and a final analysis. The first interim analysis was planned analysis, no further formal testing of progression-free to occur approximately 9 months after the last participant survival was planned. A hierarchical testing strategy was was randomised, after approximately 331 progression-free used, with progression-free survival tested first in the survival events had occurred in the PD-L1 CPS 1 or higher PD-L1 CPS 1 or higher population and then in the overall population and approximately 460 progression-free population, followed by overall survival in the same survival events had occurred in the overall population. sequence; if progression-free survival was significant, the 867 patients screened for eligibility 224 excluded* 122 did not meet inclusion criteria 84 met exclusion criteria 32 refused to participate 643 randomly assigned 322 assigned to pembrolizumab 321 assigned to placebo 2 excluded (randomised in error—not 3 excluded (randomised in error—not eligible) eligible) 320 received pembrolizumab plus chemotherapy 318 received placebo plus chemotherapy 306 discontinued treatment 309 discontinued treatment 211 progressive disease 230 progressive disease 43 adverse event 31 adverse event 31 clinical progression 33 clinical progression 18 withdrawal of consent 14 withdrawal of consent 2 physician decision 1 complete response 1 complete response 7 completed all study treatment 2 completed all study treatment 7 treatment ongoing 7 treatment ongoing 322 included in intention-to-treat population 321 included in intention-to-treat population 320 included in treatment-as-assigned population 318 included in treatment-as-assigned population Figure 1: Trial profile *Participants might have failed screening for more than one reason. The reason for failing screening was not recorded for three participants. Articles α was carried forward to overall survival. The graphical method of Mauer and Bretz was used to control the Pembrolizumab Placebo plus family-wise type I error rate at a one-sided α=0·025, plus paclitaxel paclitaxel with 0·02 initially allocated to progression-free survival in (n=322) (n=321) participants with PD-L1 CPS 1 or higher tumours and Age, years 0·005 allocated to progression-free survival in all Median (IQR) 62 (53–69) 61 (53–68) participants. Further details about sample size, power, ≥65 122 (38%) 114 (36%) testing hierarchy, alpha allocation, and interim analyses Race are provided in the appendix (pp 9–10). The study was Asian 72 (22%) 58 (18%) considered positive if progression-free survival was Black or African American 8 (2%) 6 (2%) significantly improved in the PD-L1 CPS 1 or higher Multiple 12 (4%) 17 (5%) population or the overall population according to the Native Hawaiian or Other Pacific 1 (<1%) 1 (<1%) prespecified graphical testing strategy. Islander Safety was assessed in all randomly assigned participants White 207 (64%) 217 (68%) who had received at least one dose of study treatment (the Missing 22 (7%) 22 (7%) as-treated population); adverse events were summarised ECOG performance status score* descriptively without formal between-group testing. 0 179 (56%) 175 (55%) Patient-reported outcomes were assessed in all randomly 1 142 (44%) 144 (45%) assigned participants who received at least one dose of Missing 1 (<1%) 2 (1%) study treatment and had at least one assessment for the Bevacizumab use instrument. Patient-reported outcome instrument Planned 235 (73%) 236 (74%) completion rates were calculated at each timepoint for all Actual 235 (73%) 236 (74%) randomly assigned participants; compliance rates were PD-L1 combined positive score calculated at each timepoint for the number of participants <1 88 (27%) 89 (28%) who were alive and expected to complete the questionnaire 1 to <10 133 (41%) 132 (41%) at the specified timepoint. The full statistical analysis plan ≥10 101 (31%) 100 (31%) is available in the appendix (pp 298–317). FIGO 2014 stage at diagnosis IA to IIB 24 (7%) 26 (8%) Role of the funding source IIIA to IIIC 183 (57%) 189 (59%) The funder of the study participated in study design, data IVA to IVB 115 (36%) 106 (33%) collection, data analysis, data interpretation, and writing Histology of the report. High-grade serous 278 (86%) 275 (86%) Results Clear cell 24 (7%) 26 (8%) Endometrioid 9 (3%) 4 (1%) Between Dec 13, 2021, and July 3, 2023, 643 participants were enrolled and randomly assigned to pembrolizumab Low-grade serous 6 (2%) 10 (3%) plus paclitaxel (n=322) or placebo plus paclitaxel (n=321) Carcinosarcoma 3 (1%) 5 (2%) in the intention-to-treat population. A total of Other 2 (1%) 1 (<1%) 638 participants received at least one dose of study Prior lines of therapy treatment (320 in the pembrolizumab plus paclitaxel One line 121 (38%) 113 (35%) group and 318 in the placebo plus paclitaxel group) and Two lines 200 (62%) 207 (64%) were included in the safety (as-treated) population Three lines 1 (<1%) 1 (<1%) (figure 1). Data cutoff dates were April 3, 2024, for the Prior exposure first interim analysis (median follow-up 15·6 months Anti-PD-1 or PD-L1 7 (2%) 7 (2%) [IQR 12·4–19·3]), March 5, 2025, for the second interim Bevacizumab 149 (46%) 146 (45%) analysis (26·6 months [23·4–30·4]), and Sept 5, 2025, for PARP inhibitor 112 (35%) 123 (38%) the final analysis (32·7 months [29·5–36·4]). Platinum-free interval, months <3 137 (43%) 161 (50%) 3 to 6 183 (57%) 155 (48%) >6 2 (1%) 5 (2%) Figure 2: Progression-free survival (A) Kaplan–Meier estimates of progression-free survival in the PD-L1 CPS of 1 or Data are n (%) or median (IQR). ECOG=Eastern Cooperative Oncology Group. higher population at the first interim analysis. (B) Kaplan–Meier estimates of FIGO=International Federation of Gynecology and Obstetrics. PARP=poly(ADP- progression-free survival in the overall population at the first interim analysis. ribose) polymerase. *ECOG performance status scores range from 0 to 5, with 0 indicating no symptoms and higher scores indicating greater disability. (C) Analysis of progression-free survival in protocol-specified subgroups of the PD-L1 CPS of 1 or higher population. (D) Analysis of progression-free survival in Table 1: Baseline demographic and disease characteristics in the protocol-specified subgroups of the overall population. CPS=combined positive intention-to-treat population score. ECOG=Eastern Cooperative Oncology Group. PARP=poly(ADP-ribose) polymerase. 1530 Articles BB 90 70 50 30 Hazard ratio 0·70 10 (95% CI 0·58–0·84) p<0·0001 (α=0·0023) 0 5 10 15 20 25 30 0 5 10 15 20 25 30 Time (months) Time (months) Number at risk (censored) Pembrolizumab– 322 (0) 233 (11) 119 (23) 34 (67) 6 (86) 1 (90) 0 (91) paclitaxel group Placebo– 321 (0) 200 (14) 84 (28) 19 (57) 3 (66) 1 (67) 0 (68) paclitaxel group 0·1 1 10 0·1 1 10 Pembrolizumab–paclitaxel better Placebo–paclitaxel better Pembrolizumab–paclitaxel better Placebo–paclitaxel better )%( lavivrus eerf-noissergorP AA Pembrolizumab–paclitaxel group Placebo–paclitaxel group Hazard ratio 0·72 (95% CI 0·58–0·89) p=0·0014 (α=0·012) 234 (0) 170 (10) 87 (18) 21 (56) 5 (68) 1 (71) 0 (72) 232 (0) 150 (9) 64 (21) 16 (42) 3 (50) 1 (51) 0 (52) C Number of events/ Hazard ratio D Number of events/ Hazard ratio number of (95% CI) number of (95% CI) participants (%) participants (%) Age, years Age, years <65 218/289 (75%) 0·72 (0·55–0·93) <65 311/407 (76%) 0·69 (0·55–0·87) ≥65 124/177 (70%) 0·85 (0·60–1·22) ≥65 173/236 (73%) 0·82 (0·61–1·11) Race Race White 227/313 (73%) 0·80 (0·61–1·03) White 317/424 (75%) 0·79 (0·64–0·99) All others 88/118 (75%) 0·61 (0·40–0·93) All others 131/175 (75%) 0·56 (0·40–0·80) Region Region USA 33/50 (66%) 0·68 (0·34–1·36) USA 40/59 (68%) 0·74 (0·39–1·38) EU 163/218 (75%) 0·89 (0·66–1·22) EU 230/300 (77%) 0·87 (0·67–1·13) Rest of world 146/198 (74%) 0·64 (0·46–0·89) Rest of world 214/284 (75%) 0·62 (0·47–0·81) ECOG performance status ECOG performance status 0 186/258 (72%) 0·88 (0·66–1·17) 0 266/354 (75%) 0·81 (0·64–1·03) 1 154/206 (75%) 0·64 (0·47–0·89) 1 215/286 (75%) 0·65 (0·49–0·85) Previous PARP inhibitors Previous PARP inhibitors Yes 126/179 (70%) 0·70 (0·49–0·99) Yes 172/234 (74%) 0·74 (0·55–1·00) No 216/287 (75%) 0·79 (0·61–1·04) No 312/409 (76%) 0·74 (0·59–0·92) PD-L1 CPS PD-L1 CPS 1 to <10 202/266 (76%) 0·74 (0·56–0·97) <1 142/177 (80%) 0·66 (0·47–0·92) ≥10 140/200 (70%) 0·79 (0·57–1·10) 1 to <10 202/266 (76%) 0·74 (0·56–0·97) Bevacizumab use ≥10 140/200 (70%) 0·79 (0·57–1·10) Yes 228/338 (68%) 0·75 (0·58–0·97) Bevacizumab use No 114/128 (89%) 0·71 (0·49–1·03) Yes 329/471 (70%) 0·72 (0·58–0·89) Platinum-free interval, months No 155/172 (90%) 0·70 (0·51–0·97) <3 162/220 (74%) 0·85 (0·62–1·15) Platinum-free interval, months 3 to 6 174/240 (73%) 0·71 (0·53–0·96) <3 222/299 (74%) 0·80 (0·62–1·05) Overall 342/466 (73%) 0·72 (0·58–0·89) 3 to 6 252/334 (75%) 0·71 (0·55–0·90) Overall 484/643 (75%) 0·70 (0·58–0·84) Articles BB 90 70 50 30 Hazard ratio 0·82 10 (95% CI 0·69-0·97) p=0·011 (α=0·024) 0 6 12 18 24 30 36 42 0 6 12 18 24 30 36 42 48 Time (months) Time (months) Number at risk (censored) Pembrolizumab– 322 (0) 277 (1) 211 (2) 157 (2) 105 (2) 46 (37) 15 (61) 1 (74) 0 (75) paclitaxel group Placebo– 321 (0) 271 (2) 191 (2) 124 (3) 86 (3) 36 (28) 8 (49) 1 (54) 0 (55) paclitaxel group 1532 )%( ,lavivrus llarevO AA Pembrolizumab–paclitaxel group Placebo–paclitaxel group Hazard ratio 0·76 (95% CI 0·61-0·94) p=0·0053 (α=0·0083) 234 (0) 207 (1) 161 (1) 120 (1) 49 (39) 13 (65) 3 (74) 0 (77) 232 (0) 200 (1) 137 (1) 89 (2) 41 (26) 10 (48) 1 (56) 0 (57) C Number of events/ Hazard ratio D Number of events/ Hazard ratio number of (95% CI) number of (95% CI) participants (%) participants (%) Age, years Age, years <65 206/289 (71%) 0·73 (0·55–0·96) <65 323/407 (79%) 0·74 (0·59-0·92) ≥65 126/177 (71%) 0·82 (0·58–1·16) ≥65 190/236 (81%) 0·94 (0·71-1·25) Race Race White 213/313 (74%) 0·78 (0·60–1·02) White 337/424 (80%) 0·86 (0·70-1·07) All others 90/118 (76%) 0·70 (0·46–1·06) All others 137/175 (78%) 0·69 (0·49-0·97) Region Region USA 35/50 (70%) 0·82 (0·42–1·59) USA 46/59 (78%) 0·88 (0·49-1·57) EU 148/218 (68%) 0·90 (0·65–1·24) EU 239/300 (80%) 1·02 (0·79-1·31) Rest of world 149/198 (75%) 0·63 (0·46–0·88) Rest of world 228/284 (80%) 0·63 (0·49-0·82) ECOG performance status ECOG performance status 0 170/258 (66%) 0·83 (0·61–1·12) 0 268/354 (76%) 0·85 (0·67-1·08) 1 160/206 (78%) 0·72 (0·53–0·99) 1 242/286 (85%) 0·77 (0·60-0·99) Previous PARP inhibitors Previous PARP inhibitors Yes 124/181 (69%) 0·85 (0·59–1·20) Yes 186/235 (79%) 0·91 (0·68-1·21) No 208/285 (73%) 0·71 (0·54–0·93) No 327/408 (80%) 0·75 (0·61-0·94) PD-L1 CPS PD-L1 CPS 1 to <10 203/265 (77%) 0·79 (0·60–1·04) <1 147/177 (83%) 0·98 (0·71-1·36) ≥10 129/201 (63%) 0·72 (0·51–1·01) 1 to <10 223/265 (84%) 0·76 (0·59-0·99) Bevacizumab use ≥10 143/201 (71%) 0·75 (0·54-1·04) Yes 229/338 (68%) 0·75 (0·58–0·97) Bevacizumab use No 103/128 (81%) 0·75 (0·51–1·11) Yes 357/471 (76%) 0·80 (0·65-0·99) Platinum-free interval, months No 156/172 (91%) 0·80 (0·59-1·10) <3 162/218 (74%) 0·64 (0·47–0·88) Platinum-free interval, months 3 to 6 167/245 (68%) 0·88 (0·65–1·19) <3 240/298 (81%) 0·69 (0·53-0·89) Overall 332/466 (71%) 0·76 (0·61–0·94) 3 to 6 266/338 (79%) 0·94 (0·73-1·19) Overall 513/643 (80%) 0·82 (0·69-0·97) 0·1 1 10 0·1 1 10 Pembrolizumab–paclitaxel better Placebo–paclitaxel better Pembrolizumab–paclitaxel better Placebo–paclitaxel better Articles Baseline demographic and disease characteristics were At the final analysis, overall survival was significantly well balanced between groups (table 1). Most participants improved in the overall population (median 17·7 months had PD-L1 CPS 1 or higher tumours (466 [72%] of 643) vs 14·0 months; HR 0·82, 95% CI 0·69–0·97; p=0·011 and high-grade serous carcinoma (553 [86%] of 643). [α=0·024]) and favoured pembrolizumab plus paclitaxel Bevacizumab use was planned pre-randomisation in in most subgroups (figure 3). The overall survival benefit 471 (73%) of 643 participants, and adherence to the in the PD-L1 CPS 1 or higher population continued bevacizumab stratification assignment was high. Of (HR 0·76, 95% CI 0·62–0·93) at the final analysis 643 participants, 407 (63%) had received two previous (appendix p 12). Progression-free survival results at the lines of therapy, 295 (46%) had received bevacizumab, final analysis are provided in the appendix (p 13) and are 235 (37%) had received a poly(ADP-ribose) polymerase presented descriptively, as formal hypothesis testing for (PARP) inhibitor, and 14 (2%) had received anti-PD-1/ progression-free survival had already been completed at PD-L1 therapy. The time from last platinum therapy to the first interim analysis. progression was less than 3 months in 298 (46%) and Investigator-assessed objective response rates were 3–6 months in 338 (53%) of 643 participants. higher with pembrolizumab plus paclitaxel than with At the first interim analysis, pembrolizumab plus placebo plus paclitaxel in both the PD-L1 CPS 1 or higher paclitaxel significantly improved progression-free survival population (53% [107 of 202] vs 47% [95 of 204]; versus placebo plus paclitaxel in both the PD-L1 CPS 1 or difference 5·8, 95% CI –3·6 to 15·1) and the overall higher (median 8·3 months vs 7·2 months; HR 0·72, population (50% [139 of 276] vs 41% [116 of 282]; 95% CI 0·58–0·89; p=0·0014 [α=0·012]) and overall difference 8·5, 95% CI 0·6 to 16·4; appendix p 17). The populations (8·3 months vs 6·4 months; 0·70, 0·58–0·84; response durations were longer with pembrolizumab p<0·0001 [α=0·0023]; figure 2), meeting the prespecified plus paclitaxel (10·4 months vs 8·1 months in PD-L1 statistical criteria for these primary hypotheses. Because CPS 1 or higher population; 10·4 months vs 7·9 months the prespecified efficacy boundary was crossed at this in overall population) and a greater proportion of analysis, the progression-free survival hypotheses were participants had a response duration of minimum considered formally rejected and no further α-controlled 12 months (47% [47 of 107] vs 30% [26 of 95] in the PD-L1 testing of progression-free survival was performed at CPS 1 or higher population; 47% [60 of 139] vs 28% subsequent analyses. Progression-free survival was [30 of 116] in overall population; appendix p 14) generally consistent across prespecified subgroups in both In the as-treated population, the median duration of populations. Progression-free survival, as assessed by treatment was 32·5 weeks (IQR 18·4–60·6) in the blinded independent central review, was concordant with pembrolizumab plus paclitaxel group and 27·6 weeks investigator-assessed results (appendix p 16). Overall (16·1–46·1) in the placebo plus paclitaxel group; the survival was not significantly improved at the first interim median number of cycles was six in the pembrolizumab analysis in the PD-L1 CPS 1 or higher population (HR 0·72, plus paclitaxel group and five in the placebo plus 95% CI 0·55–0·94; p=0·0084 [α=0·0021]), and the study paclitaxel group (appendix p 18). Adverse events of any proceeded to the second interim analysis. cause occurred in 319 (100%) of 320 participants in the At the second interim analysis, overall survival was pembrolizumab plus paclitaxel group and in 316 (99%) of significantly improved in the PD-L1 CPS 1 or higher 318 participants in the placebo plus paclitaxel group; population (median 18·2 months vs 14·0 months; grade 3 or higher adverse events occurred in 265 (83%) of HR 0·76, 95% CI 0·61–0·94; p=0·0053 [α=0·0083]) and 320 participants in the pembrolizumab plus paclitaxel favoured pembrolizumab plus paclitaxel in all subgroups group and 225 (71%) of 318 participants in the placebo (figure 3). For overall survival in the overall population, plus paclitaxel group (table 2). the HR favoured the pembrolizumab plus paclitaxel Grade 3 or higher treatment-related adverse events group, although statistical significance was not reached occurred in 217 (68%) of 320 participants in the (HR 0·81, 95% CI 0·68–0·97; p=0·011 [α=0·0084]). The pembrolizumab plus paclitaxel group and in 176 (55%) of progression-free survival benefit with pembrolizumab 318 participants in the placebo plus paclitaxel group. The plus paclitaxel continued in the PD-L1 CPS 1 or higher most common treatment-related adverse events in both population (HR 0·75 [95% CI 0·61–0·91]) and the overall groups were anaemia, peripheral neuropathy, alopecia, population (HR 0·73 [95% CI 0·62–0·86]; appendix p 11). fatigue, and nausea. Serious treatment-related adverse events occurred in 107 (33%) of 320 participants in the pembrolizumab plus paclitaxel group and 63 (20%) of 318 participants in the placebo plus paclitaxel group. Figure 3: Overall survival Treatment-related adverse events led to discontinuation (A) Kaplan–Meier estimates of overall survival in the PD-L1 CPS of 1 or higher population at the second interim analysis. (B) Kaplan-Meier estimates of overall of any trial treatment in 120 (38%) participants in the survival in the overall population at the final analysis. (C) Analysis of overall pembrolizumab plus paclitaxel group and in 90 (28%) survival in protocol-specified subgroups of the PD-L1 CPS of 1 or higher participants in the placebo plus paclitaxel group. In the population. (D) Analysis of overall survival in protocol-specified subgroups of two groups, respectively, 37 (12%) and 20 (6%) participants the overall population. CPS=combined positive score. ECOG=Eastern Cooperative Oncology Group. PARP=poly(ADP-ribose) polymerase. discontinued pembrolizumab or placebo, 91 (28%) and Articles in 38 (12%) of 320 participants in the pembrolizumab Pembrolizumab plus Placebo plus paclitaxel plus paclitaxel group and in 11 (3%) of 318 participants in paclitaxel (n=320) (n=318) the placebo plus paclitaxel group. Deaths due to immune- Any grade Grade ≥3 Any grade Grade ≥3 mediated adverse events occurred in two (1%) participants Any adverse event* 319 (100%) 265 (83%) 316 (99%) 225 (71%) in the pembrolizumab plus paclitaxel group (colitis and Any treatment-related adverse event† 313 (98%) 217 (68%) 303 (95%) 176 (55%) pneumonitis) and in none of the participants in the Serious adverse event‡ 180 (56%) ·· 123 (39%) ·· placebo plus paclitaxel group. Serious treatment-related adverse event† 107 (33%) ·· 63 (20%) ·· At week 24, the completion rate for the EORTC Adverse event leading to discontinuation 136 (43%) ·· 109 (34%) ·· QLQ-C30 was 69% (216 of 311 participants) in the Treatment-related adverse event leading to 120 (38%) ·· 90 (28%) ·· pembrolizumab plus paclitaxel group and 58% (184 of discontinuation 315 participants) in the placebo plus paclitaxel group Adverse event leading to death 16 (5%) ·· 14 (4%) ·· (compliance rates were 91% [216 of 237 participants] and Treatment-related adverse event leading to death 4 (1%) ·· 5 (2%) ·· 86% [184 of 214 participants], respectively). Global health Treatment-related adverse events occurring in ≥20% of participants in a trial group status or quality of life improved (≥10 point increase Anaemia 159 (50%) 38 (12%) 134 (42%) 25 (8%) from baseline) or remained stable in 183 (59%) of Peripheral neuropathy 124 (39%) 17 (5%) 99 (31%) 19 (6%) 311 participants in the pembrolizumab plus paclitaxel Alopecia 121 (38%) 0 108 (34%) 0 group and 193 (61%) of 315 participants in the placebo Fatigue 113 (35%) 17 (5%) 105 (33%) 19 (6%) plus paclitaxel group. Comparable results between Nausea 100 (31%) 4 (1%) 87 (27%) 4 (1%) patient-reported outcome assessment instruments were Epistaxis 94 (29%) 1 (<1%) 69 (22%) 0 observed for the abdominal and gastrointestinal Diarrhoea 93 (29%) 10 (3%) 79 (25%) 3 (1%) symptom scale of the EORTC QLQ-OV28 (229 [74%] of Neutrophil count decreased 92 (29%) 59 (18%) 71 (22%) 43 (14%) 311 participants and 215 [68%] of 315 participants, White blood cell count decreased 75 (23%) 38 (12%) 57 (18%) 28 (9%) respectively). Neutropenia 73 (23%) 45 (14%) 78 (25%) 44 (14%) Immune-mediated adverse event§ 126 (39%) 38 (12%) 60 (19%) 11 (3%) Discussion Hypothyroidism 58 (18%) 0 19 (1%) 0 In the phase 3 ENGOT-ov65/KEYNOTE-B96 trial, Infusion reaction 19 (6%) 6 (2%) 15 (5%) 2 (1%) pembrolizumab combined with weekly paclitaxel, with or Hyperthyroidism 16 (5%) 0 2 (1%) 0 without bevacizumab, significantly improved progression- Adrenal insufficiency 15 (5%) 7 (2%) 0 0 free survival and overall survival in participants with Pneumonitis 15 (5%) 2 (1%) 6 (2%) 3 (1%) platinum-resistant recurrent ovarian cancer who had received one to two previous systemic regimens. The *Listed are adverse events that occurred during the treatment period or within 30 days after the treatment period (within 90 days for serious events). The as-treated population included all the participants who had undergone magnitude of benefit was clinically meaningful and randomisation and received at least one dose of study treatment. The severity of adverse events was graded according observed in both the PD-L1 CPS 1 or higher population to the Common Terminology Criteria for Adverse Events, version 5.0, of the National Cancer Institute. †Treatment- and the overall population. Although PD-L1 expression in related adverse events were events that were attributed to the trial treatment by the investigators; treatment-related ovarian cancer is heterogeneous, previous studies of PD-1 adverse events that occurred in at least 20% of the participants are reported; the events are listed in descending order of frequency in the pembrolizumab–paclitaxel group; participants can have had more than one event. ‡Defined as any blockade have suggested greater activity in PD-L1–positive untoward medical occurrence that was life-threatening, required hospitalisation, or resulted in disability or death. tumours. In this trial, the improvement in overall survival §Immune-mediated adverse events were determined according to a list of terms specified by the sponsor, regardless of was greater in the CPS 1 or higher population than in the attribution to any trial treatment by the investigators; adverse events of interest that occurred in at least 15 participants are reported. overall population, supporting its use as an enrichment strategy. Notably, HRs consistently favoured Table 2: Adverse events in either treatment group of the as-treated population pembrolizumab across CPS strata in the overall population, supporting the robustness of these results. These findings 78 (25%) participants discontinued paclitaxel, and 46 underscore the complexity of PD-L1 as a predictive (14%) and 33 (10%) participants discontinued biomarker in ovarian cancer and the contribution of bevacizumab due to treatment-related adverse events. additional tumour and host immune factors beyond PD-L1 Treatment-related adverse events resulted in death in expression alone, highlighting the need for more refined four participants (1%) in the pembrolizumab plus biomarkers to better characterise immune responsiveness paclitaxel group (colitis, interstitial lung disease, acute in this disease. The benefit of pembrolizumab was myeloid leukaemia, and intestinal perforation) and in generally consistent across prespecified subgroups, five participants (2%) in the placebo plus paclitaxel group including participants with adverse prognostic features (cardiac failure, intestinal perforation [in two participants], such as older age, previous exposure to PARP inhibitors, and large-intestine perforation [in two participants]). and short platinum-free intervals. Benefits were observed Potentially immune-mediated adverse events or in both participants who received bevacizumab and those infusion reactions occurred in 126 (39%) of who did not, supporting the robustness of the treatment 320 participants in the pembrolizumab plus paclitaxel effect across contemporary treatment patterns. group and in 60 (19%) of 318 participants in the placebo These results differ from those of previous phase 3 trials plus paclitaxel group; grade 3 or higher events occurred evaluating the PD-L1 inhibitors atezolizumab and 1534 Articles avelumab in combination with single-agent chemotherapy reduced paclitaxel exposure. Most discontinuations owing in platinum-resistant ovarian cancer, which did not show to treatment-related adverse events involved paclitaxel, significant survival benefits.10,11 One plausible explanation which was continued per protocol until disease for this difference is the choice of chemotherapy backbone. progression or any other protocol-specified reason to Weekly (metronomic) paclitaxel has immunomodulatory discontinue treatment. The median number of paclitaxel properties that might enhance antitumour immune administrations was 20·5 (22 in the pembrolizumab responses, including modulation of antigen presentation group and 19 in the placebo group), supporting the and immunos uppressive cell populations.12 Consistent interpretation that these discontinuations reflect with this hypothesis, exploratory analyses from the cumulative adverse effects rather than additive toxicity atezolizumab trial suggested greater benefit when from pembrolizumab. Immune-mediated adverse events immune checkpoint inhibition was combined with weekly occurred more frequently in the pembrolizumab group paclitaxel than with pegylated liposomal doxorubicin.10 and were managed according to established clinical Differences in trial design, participant selection, and algorithms (eg, corticosteroids and treatment inhibition of the PD-1 versus PD-L1 pathway might also interruption) and no new safety signals were identified. have contributed to the divergent outcomes across studies. Patient-reported outcomes remained generally stable over Following decades of predominantly negative phase 3 time in both treatment groups. trials in platinum-resistant ovarian cancer, several recent A key strength of this study is the use of a control studies have reported clinically meaningful improvem ents regimen that reflects the most active standard-of-care in survival outcomes. The MIRASOL trial showed approach for platinum-resistant ovarian cancer—weekly improved progression-free survival and overall survival paclitaxel with or without bevacizumab5,23—thereby with the folate receptor α-targeted antibody–drug providing a clinically relevant benchmark. One limitation conjugate mirvetuximab soravtansine in patients with is the investigator-selected use of bevacizumab, which folate receptor α-high disease,17,18 supporting its regulatory may have been influenced by patient-related or disease- approval for this biomarker-defined population. More related characteristics. Despite this potential selection recently, the SCORES trial showed that adding the VEGF bias, treatment effects were generally consistent across antibody suvemcitug to chemotherapy in platinum- clinical endpoints in participants treated with or without resistant ovarian cancer significantly improved bevacizumab. Nonetheless, the study was not designed progression-free survival and overall survival compared or powered to determine the independent contribution with chemotherapy alone, although the magnitude of of bevacizumab to the treatment effect, and the specific overall survival benefit was modest.6 In the ROSELLA role of bevacizumab within this combination cannot be trial, the addition of relacorilant to nab-paclitaxel definitively established from these data. Additionally, significantly prolonged progression-free survival19 and eligibility was restricted to participants who had received overall survival.20 Together, these findings suggest that no more than two previous lines of systemic therapy, biologically selected or targeted strategies can improve which could limit generalisability to more heavily outcomes in this historically treatment-refractory pretreated populations. Finally, the rate of prior PARP population. ENGOT-ov65/KEYNOTE-B96 builds on this inhibitor exposure reflects variability in global access and evolving landscape as the first phase 3 study to show an evolving clinical practice during the study period and overall survival benefit with an immune checkpoint evolving biomarker-driven treatment strategies. inhibitor-based regimen in platinum-resistant ovarian In conclusion, pembrolizumab in combination with cancer. Based on these data, pembrolizumab plus weekly paclitaxel, with or without bevacizumab, resulted paclitaxel, with or without bevacizumab, received in statistically significant and clinically meaningful regulatory approval for patients with platinum-resistant improvements in progression-free survival and overall ovarian cancer whose tumours express PD-L1 with a CPS 1 survival in participants with platinum-resistant recurrent or higher and who have received one to two previous ovarian cancer. These findings provide randomised systemic regimens,21,22 and is now a recommended option phase 3 evidence supporting this regimen as an effective in National Comprehensive Cancer Network guidelines,4 treatment option in this setting. establishing immunoc hemotherapy as a standard Contributors treatment in this setting. An important consideration for All authors participated in the design and conduct of the study; or the clinical practice is how to position immunochemotherapy collection, management, analysis, and interpretation of the data. All authors participated in the preparation, review, or approval of the relative to other agents; in the absence of head-to-head manuscript; and the decision to submit the manuscript for publication. data, treatment selection should be guided by biomarker NC and EZ wrote the first draft of the manuscript with medical writing status, previous therapies, toxicity profiles, and clinical assistance and with input from XP, KY, and AMB. NC, XP, KY, and AMB goals. directly accessed and verified the data and statistical analyses. The safety profile of pembrolizumab plus weekly Declaration of interests paclitaxel was consistent with the known safety profiles of NC reports grants or contracts from AstraZeneca and GSK (paid to institution), payment or honoraria for lectures, presentations, speakers the individual agents. Chemotherapy-associated adverse bureaus, manuscript writing, or educational events from AstraZeneca, events predominated, and there was no evidence of MSD, GSK, Eisai, AbbVie, Pharma&, and Corcept; support for attending Articles meetings or travel from GSK, AstraZeneca, and AbbVie; participation on Oncology, Pan-Arabian Research Society for Gynecological Oncology, a Data Safety Monitoring Board or Advisory Board from Roche, Deutsch-Türkische Gynäkologengesellschaft, Deutsche Stiftung AstraZeneca, MSD, GSK, Immunogen, Eisai, Oncxerna, Nuvation Bio, Eierstockkrebs, Arbeitsgemeinschaft Gynäkologische Onkologie, and Gilead, Regeneron, Novocure, Seagen, AbbVie, Lilly, BeOne, and Charité Comprehensive Cancer Center; medical writing assistance from Biontech; medical writing assistance from MSD; and participation in the MSD; unpaid participation in editorial boards for International Journal of Scientific Committee ACTO and the Member Nomination Committee Gynecological Cancer, Journal of the Turkish-German Gynecological ESMO (unpaid). EZ reports grants or contracts from MSD, NCI, DoD, Association, Anticancer Research, Archives of Gyencology and Obstetrics, and OCRA; consulting fees from MSD, AbbVie, NextCure, Iovance, Research and Treatment, Annals of Gynecology and Obstetrics Research, 92Biotech, Takeda Oncology, and Apellis; payment or honoraria for Journal of Hematology & Multiple Myeloma, and guest editorship for lectures, presentations, speakers bureaus, manuscript writing, or Cancers. M-EP reports payment or honoraria for lectures, presentations, educational events for AstraZeneca; and participation on a Data Safety speakers bureaus, manuscript writing, or educational events from GSK, Monitoring Board or Advisory Board for MSD. ER reports grants or AstraZeneca, AbbVie, Tecnofarma, Dr Reddy’s, and MSD; and support contracts from AstraZeneca, Gilead, MSD, and Roche (paid to for attending meetings or travel from GSK, AstraZeneca, Roche, AbbVie, institution); and support for attending meetings or travel and and MSD. SH reports participation on a Data Safety Monitoring Board or participation in a Data Monitoring Committee from Philogen. AS reports Advisory Board and advisory or consulting fees from AstraZeneca, Eisai, Advisory Board and Data Safety Monitoring Board participation, and MSD, GSK, and AbbVie. CZ reports honoraria for participation in the speaker’s honorarium, from GSK, Eisai, and MSD. CG reports Delphi study from Daiichi Sankyo and AstraZeneca; payment or consulting fees from GSK, Gilead, MSD, and Adium; payment or honoraria for lectures, presentations, speakers bureaus, manuscript honoraria for lectures, presentations, speakers bureaus, manuscript writing, or educational events from Roche, Novartis, Pfizer, Lilly, writing, or educational events from GSK, MSD, and Adium; and AstraZeneca, Daiichi Sankyo, MSD, Eisai, GSK, Menarini Stemline, participation on a Data Safety Monitoring Board or Advisory Board from Exact Sciences, and Gilead; support for attending meetings or travel from GSK, Gilead, and Adium. KH reports research contracts from Daiichi Roche, Novartis, Pfizer, AstraZeneca, Daiichi Sankyo, MSD, Eisai, GSK, Sankyo, Eisai, MSD, and Takeda; consulting fees from Chugai, Fuji, and and Gilead; and participation on a Data Safety Monitoring Board or Kyowa Kirin; advisory board fees from Eisai, Takeda, MSD, Regeneron, Advisory Board for Roche, Novartis, Pfizer, Lilly, AstraZeneca, Daiichi Roche, Genmab, Sanofi, GSK, and Zymeworks; honoraria from Daiichi Sankyo, MSD, Eisai, GSK, Menarini Stemline, Exact Sciences, and Sankyo, AstraZeneca, Chugai, Eisai, Fuji, Genmab, MSD, Takeda, Sanofi, Gilead. BJM reports consulting fees from AstraZeneca, BioNtech, Kyowa Kirin, Kaken, GSK, and Regeneron. MM-M reports research Corcept, DSI, Eisai, Eli Lilly, Genmab/Seagen/Pfizer, GOG Foundation, funding from MSD; grants or contracts from Eli Lilly, MSD, and Fortrea GSK, Immunogen/Abbvie, Incyte, Karyopharm, MSD, Mersana, Mural/ (paid to institution); Advisory Board and speaker fees from Eli-Lilly, Alkermes, Myriad, Natera, Novartis, Novocure, OncoC4, Panavance, Roche, Merck GA, MSD, and GSK; and support for attending meetings Pharma&, ProfoundBio/Genmab, Regeneron, Roche/Genentech, Sutro, or travel from Novartis, Merck GA, and Servier. JRK reports grants or Tubulis, Verastem, Zentalis, and Zymeworks; and payment or honoraria contracts from AstraZeneca, Novartis, Philips, Dutch Cancer Society, and for lectures, presentations, speakers bureaus, manuscript writing, or Bontius Stichting/LUF (paid to institution); consulting fees from educational events from AstraZeneca, DSI, Eisai, GSK, Immunogen/ AstraZeneca, Eisai, GSK, Lilly, MSD, Novartis, and Daiichi Sankyo (paid AbbVie, and MSD. RLC reports payment or honoraria for lectures, to institution); payment or honoraria for lectures, presentations, speakers presentations, speakers bureaus, manuscript writing, or educational bureaus, manuscript writing, or educational events from Eisai, GSK, and events from AstraZeneca, MSD, GSK, Karyopharm, Faeth, Genmab, and MSD (paid to institution); support for attending meetings or travel from DSI; participation on a Data Safety Monitoring Board or Advisory Board Daiichi Sankyo (paid to institution); and participation on a Data Safety for AstraZeneca, Genmab/Seagen, GSK, Immunogen, Panavance, Monitoring Board or Advisory Board for DSMB, TEIPP, and Alison. Profoundbio, and Roche/Genentech; and leadership or fiduciary role in ACdM reports grants or contracts, paid to institution, from Amgen, other board, society, committee or advocacy group, paid or unpaid, as a AstraZeneca, Bristol Myers, Squibb, Clovis Oncology, GSK, MSD, steering committee member for AstraZeneca, GSK, MSD, Roche- Novartis, Regeneron, and Roche; payment or honoraria for lectures, Genentech, and Verastem; as a trial chair for Immonogen/AbbVie; as a presentations, speakers bureaus, manuscript writing, or educational coordinating principal investigator for Karyopharm; and as a principal events from AstraZeneca, Bristol Myers Squibb, GSK, MSD, Novartis, investigator for AbbVie, Genmab, Immunogen, MSD, and Roche/ Adium, Daiichi, Pfizer, and AbbVie; support for attending meetings or Genentech. TDLMR reports grants or contracts from GSK (paid to travel from MSD, Daiichi, and AbbVie; and participation on a Data Safety institution); consulting fees from GSK, AstraZeneca, Eisai, MSD, Pfizer, Monitoring Board or Advisory Board from AstraZeneca, Bristol Myers Novartis, Gilead, AbbVie, Sanofi, and Regeneron; payment or honoraria Squibb, GSK, MSD, Novartis, Roche, and Daiichi. PRD reports for lectures, presentations, speakers bureaus, manuscript writing, or consulting fees from MSD, Astellas, and Ipsen; support for attending educational events from GSK, AstraZeneca, MSD, Pfizer, Gilead, and meetings or travel from Ipsen and MSD; and stock or stock options from AbbVie; support for attending meetings or travel from Roche, Gilead, and Mural Oncology and UCB. J-WK reports a leadership or fiduciary role in MSD; and participation on a Data Safety Monitoring Board or Advisory the Executive Board of the Asia-Pacific Gynecologic Cancer Trials Group. Board for MSD. XP, KY, and AMB are employees of Merck Sharp & JS reports research funding from AstraZeneca, Bayer, Clovis Oncology, Dohme, a subsidiary of Merck & Co, Rahway, NJ, USA, who may own GSK, Iqvia, Lilly, MSD, Mural, Roche Pharma, and Tesaro (paid to stock or hold stock options in Merck & Co, Rahway, NJ, USA. All other institution); consulting fees from AstraZeneca, Bayer Vital, Bristol Myers, authors declare no competing interests. Clovis Oncology, Daiichi Sankyo, Eisai, Encare, Immunogen, Intuitive Data sharing Surgical, Karyopharm Therapeutics, Merck/Pfizer, MSD, Mundipharma, Merck Sharp & Dohme, a subsidiary of Merck & Co, Rahway, NJ, USA Novocure, Oncoinvent, PharmaMar, Roche Pharma, Sanofi-Aventis (MSD) is committed to providing qualified scientific researchers access Deutschland, Seagen, and Tubulis; payment or honoraria for lectures, to anonymised data and clinical study reports from the company’s presentations, speakers bureaus, manuscript writing, or educational clinical trials for the purpose of conducting legitimate scientific research. events from AMGEN, AstraZeneca, Bayer, Clovis Oncology, Corcept MSD is also obligated to protect the rights and privacy of trial Therapeutics, Daiichi Sankyo, Eisai, Esteve Pharmaceuticals, GSK, Hexal, participants and, as such, has a procedure in place for evaluating and Incyte Biosciences, Jenapharm, Kyowa Kirin, Medtronic Covidien, fulfilling requests for sharing company clinical trial data with qualified Novartis Pharma, Oncoinvent, Pharma&, PharmaMar, Phytolife external scientific researchers. The MSD data sharing website (https:// Nutrition, Roche Pharma, and Vifor Pharma; support for attending externaldatasharing-msd.com) outlines the process and requirements meetings or travel from AstraZeneca, GSK, Roche Pharma, Novocure, for submitting a data request. Applications will be promptly assessed for Immunogen, Incyte, MSD, and Eisai; participation on a Data Safety completeness and policy compliance. Feasible requests will be reviewed Monitoring Board or Advisory Board for Immunogen, Incyte, GSK, by a committee of MSD subject matter experts to assess the scientific AstraZeneca, Clovis, Novocure, Bristol Myers Squibb, MSD, Bayer, and validity of the request and the qualifications of the requestors. In line PharmaMar; leadership or fiduciary role in European Society of with data privacy legislation, submitters of approved requests must enter Gynecological Oncology, North-Eastern German Society of Gynecological 1536 Articles into a standard data-sharing agreement with MSD before data access is 7 Richardson DL, Eskander RN, O’Malley DM. Advances in ovarian granted. Data will be made available for request after product approval in cancer care and unmet treatment needs for patients with platinum the USA and EU or after product development is discontinued. There resistance: a narrative review. JAMA Oncol 2023; 9: 851–59. are circumstances that may prevent MSD from sharing requested data, 8 Varga A, Piha-Paul S, Ott PA, et al. Pembrolizumab in patients with including country-specific or region-specific regulations. If the request is programmed death ligand 1-positive advanced ovarian cancer: declined, it will be communicated to the investigator. Access to genetic analysis of KEYNOTE-028. Gynecol Oncol 2019; 152: 243–50. or exploratory biomarker data requires a detailed, hypothesis-driven 9 Matulonis UA, Shapira-Frommer R, Santin AD, et al. Antitumor statistical analysis plan that is collaboratively developed by the requestor activity and safety of pembrolizumab in patients with advanced recurrent ovarian cancer: results from the phase II KEYNOTE-100 and MSD subject matter experts; after approval of the statistical analysis study. Ann Oncol 2019; 30: 1080–87. plan and execution of a data-sharing agreement, MSD will either 10 Harter P, Marme F, Redondo A, et al. Atezolizumab with perform the proposed analyses and share the results with the requestor bevacizumab and nonplatinum chemotherapy for recurrent ovarian or will construct biomarker covariates and add them to a file with clinical cancer: final results from the placebo-controlled AGO-OVAR 2·29/ data that is uploaded to an analysis portal so that the requestor can ENGOT-ov34 phase III trial. J Clin Oncol 2026; 44: 103–16. perform the proposed analyses. 11 Pujade-Lauraine E, Fujiwara K, Ledermann JA, et al. Avelumab Acknowledgments alone or in combination with chemotherapy versus chemotherapy Funding for this research was provided by Merck Sharp & Dohme, a alone in platinum-resistant or platinum-refractory ovarian cancer subsidiary of Merck & Co, Rahway, NJ, USA. We thank the participants (JAVELIN Ovarian 200): an open-label, three-arm, randomised, phase 3 study. Lancet Oncol 2021; 22: 1034–46. and their families and caregivers for participating in this study, along with all investigators and site personnel, and the following current or former 12 Muraro E, Vinante L, Fratta E, et al. Metronomic chemotherapy: anti-tumor pathways and combination with immune checkpoint employees of Merck Sharp & Dohme, a subsidiary of Merck & Co, inhibitors. Cancers (Basel) 2023; 15: 2471. Rahway, NJ, USA: Gursel Aktan, for research supervision; Sarper Toker, 13 Pfirschke C, Engblom C, Rickelt S, et al. Immunogenic for technical help, research supervision, and scientific expertise; chemotherapy sensitizes tumors to checkpoint blockade therapy. Ruixue Wang, for statistical expertise; Silvia Senese, for data collection, Immunity 2016; 44: 343–54. research supervision, and administrative and logistical support; 14 Vafaei S, Zekiy AO, Khanamir RA, et al. Combination therapy with Eleanor Readinger, for research supervision and technical help; immune checkpoint inhibitors (ICIs); a new frontier. Cancer Cell Int Peggy Lau, for research supervision; Pete Miles, for other contributions; 2022; 22: 2. Karyn O’Flaherty, for administrative or logistical support; Amy S Blum, 15 Wenham RM, Buras AL, Gordon SW, et al. A phase 2 study of for administrative or logistical support; Elliott Blackwood, for research pembrolizumab and weekly paclitaxel for platinum-resistant supervision; Jia Lu, for research supervision; Thomas Jemielita, for epithelial ovarian cancer. Int J Gynecol Cancer 2025; published statistical expertise; Xuan Zhou, for study design and statistical expertise; online Dec 13. https://doi.org/10.1016/j.ijgc.2025.102856. Sara Mallary, for research supervision and logistical support; 16 Zsiros E, Lynam S, Attwood KM, et al. Efficacy and safety of Nuria Mara Rodriguez Moncalvo, for data collection; Jennifer Himmel, for pembrolizumab in combination with bevacizumab and oral other contributions; Rossen Roussev, for administrative and logistical metronomic cyclophosphamide in the treatment of recurrent support and other contributions; Jennifer Kelly, for data collection; ovarian cancer: a phase 2 nonrandomized clinical trial. JAMA Oncol Feng Qiao, for data collection and data management; Yuxin He, for data 2021; 7: 78–85. collection and data management; Leslie Boyle, for clinical imaging subject 17 Matulonis UA, Lorusso D, Oaknin A, et al. Efficacy and safety of matter expertise; Gregory Goldmacher, for technical help and data mirvetuximab soravtansine in patients with platinum-resistant collection; Matthew Horn, for research supervision and other ovarian cancer with high folate receptor alpha expression: results contributions; Jacob Rotmensch, for research supervision; Venkata Meka, from the SORAYA study. J Clin Oncol 2023; 41: 2436–45. for other contributions; Jing Zhao, for research supervision and statistical 18 Moore KN, Angelergues A, Konecny GE, et al. Mirvetuximab expertise; Allison Martin Nguyen, for provision of study materials and soravtansine in FRα-positive, platinum-resistant ovarian cancer. N Engl J Med 2023; 389: 2162–74. contribution to study design and interpretation; Helen Ding, for critical review and statistical expertise; Christine McCrary Sisk, for writing or 19 Olawaiye AB, Gladieff L, O’Malley DM, et al. Relacorilant and nab- paclitaxel in patients with platinum-resistant ovarian cancer editorial assistance; and Michele McColgan, for administrative support. (ROSELLA): an open-label, randomised, controlled, phase 3 trial. References Lancet 2025; 405: 2205–16. 1 Bray F, Laversanne M, Sung H, et al. Global cancer statistics 2022: 20 Lorusso D, Gladieff L, O’Malley DM, et al. Overall survival with GLOBOCAN estimates of incidence and mortality worldwide for relacorilant and nab-paclitaxel in patients with platinum-resistant 36 cancers in 185 countries. CA Cancer J Clin 2024; 74: 229–63. ovarian cancer (ROSELLA): a phase 3 randomised controlled trial. 2 van Zyl B, Tang D, Bowden NA. Biomarkers of platinum resistance Lancet 2026; published online April 10. https://doi.org/10.1016/ in ovarian cancer: what can we use to improve treatment. S0140-6736(26)00462-9. Endocr Relat Cancer 2018; 25: R303–R18. 21 FDA. FDA approves pembrolizumab with paclitaxel for platinum- 3 Yang L, Xie HJ, Li YY, Wang X, Liu XX, Mai J. Molecular resistant epithelial ovarian, fallopian tube, or primary peritoneal mechanisms of platinum-based chemotherapy resistance in ovarian carcinoma. 2026. https://www.fda.gov/drugs/resources-information- cancer (review). Oncol Rep 2022; 47: 82. approved-drugs/fda-approves-pembrolizumab-paclitaxel-platinum- 4 National Comprehensive Cancer Network. NCCN Guidelines. resistant-epithelial-ovarian-fallopian-tube-or?utm_ Ovarian Cancer/Fallopian Tube Cancer/Primary Peritoneal Cancer. medium=email&utm_source=govdelivery (accessed March 2, 2026). Version 1. 2026. https://www.nccn.org/guidelines/guidelines- 22 European Medicines Agency. Keytruda—opinion on variation to detail?category=1&id=1453 (accessed March 2, 2026). marketing authorisation. 2026. https://www.ema.europa.eu/en/ 5 Pujade-Lauraine E, Hilpert F, Weber B, et al. Bevacizumab medicines/human/variation/keytruda-1 (accessed March 2, 2026). combined with chemotherapy for platinum-resistant recurrent 23 Osman MA, Elkady MS, Nasr KE. Weekly paclitaxel versus three- ovarian cancer: the AURELIA open-label randomized phase III trial. weekly paclitaxel in recurrent platinum-resistant epithelial ovarian J Clin Oncol 2014; 32: 1302–08. and peritoneal cancers: a phase III study. Clin Med Insights Oncol 6 Yuan G, Lou G, Li J, et al. Suvemcitug plus chemotherapy in women 2016; 10: 35–41. with platinum-resistant recurrent ovarian cancer: the SCORES randomized, double-blinded, phase 3 trial. Nat Cancer 2026; 7: 182–93. --- [PDF原文](https://sci-net.xyz/storage/7751365/d6aa6a7f3a894232a77ed21a1d7bb96b48ba143addfc6a3f329db2f2a849cd99/Pembrolizumab-plus-weekly-paclitaxel-in-platinum-resistant-recurrent-ovarian-cancer-ENGOT.pdf) DOI: 10.1016/S0140-6736(26)00602-1