BACKGROUND: A need exists for first-line chemotherapy-free options in advanced or recurrent endometrial cancer. We therefore assessed the efficacy, quality of life, and safety of first-line pembrolizumab versus carboplatin plus paclitaxel in mismatch repair-deficient (dMMR) advanced or recurrent endometrial cancer. METHODS: The KEYNOTE-C93/GOG-3064/ENGOT-en15 multicentre, open-label, randomised, active-comparator controlled, phase 3 trial was conducted at 122 centres in 24 countries in the Americas, Asia, Europe, and Oceania. Adult participants (≥18 years) with inoperable, stage III or IV, persistent, advanced, or recurrent dMMR endometrial cancer were randomly assigned 1:1 to intravenous pembrolizumab 400 mg every 6 weeks for up to 18 cycles or intravenous carboplatin area under the curve 5 mg/mL or 6 mg/mL per min plus paclitaxel 175 mg/m2 on day 1 of each 3-week cycle for six cycles. Randomisation was stratified by disease status, histology, and previous receipt of (neo)adjuvant chemotherapy. Primary endpoints were progression-free survival per RECIST version 1.1, by blinded independent central review, and overall survival. This trial is registered at ClinicalTrials.gov, NCT05173987, and is ongoing. FINDINGS: Between Feb 22, 2022, and Dec 12, 2023, 299 participants were assigned to pembrolizumab (n=149) or carboplatin plus paclitaxel (n=150). At the cutoff for the prespecified interim analysis in the intention-to-treat population (April 24, 2026), median progression-free survival was 38·2 months (95% CI 17·9 months-not reached) in the pembrolizumab group versus 8·4 months (6·8-10·4) in the carboplatin plus paclitaxel group (hazard ratio [HR] 0·57, 95% CI 0·42-0·79; p=0·0002). Estimated progression-free survival was 75% (95% CI 67-81) versus 76% (68-82) at 6 months, 67% (59-74) versus 40% (32-48) at 12 months, and 53% (44-61) versus 30% (22-38) at 36 months. The prespecified statistical significance criterion for overall survival was not met at the interim analysis (medians not reached; HR 0·77, 95% CI 0·52-1·15; p=0·10). Grade 3 or higher treatment-related adverse events occurred in 27 (18%) of 149 participants in the pembrolizumab group and 73 (51%) of 144 in the carboplatin plus paclitaxel group. Treatment-related adverse events of any grade that occurred in at least 25% of participants in either group included alopecia (three [2%] of 149 in the pembrolizumab group, 72 [50%] of 144 in the carboplatin plus paclitaxel group), anaemia (seven [5%], 71 [49%]), nausea (seven [5%], 41 [28%]), decreased neutrophil count (five [3%], 40 [28%]), peripheral neuropathy (none, 38 [26%]), and fatigue (20 [13%], 37 [26%]). Treatment-related adverse events led to treatment discontinuation in 14 (9%) of 149 participants in the pembrolizumab group and 21 (15%) of 144 participants in the carboplatin plus paclitaxel group; they resulted in death in two participants in the pembrolizumab group (hepatic failure and bacterial meningitis [n=1 each]) and one participant in the carboplatin plus paclitaxel group (pulmonary embolism). INTERPRETATION: First-line pembrolizumab showed a statistically significant and clinically meaningful improvement in progression-free survival, with lower frequency of overall and grade 3 or above treatment-related adverse events. The study did not compare pembrolizumab monotherapy with the current standard of care (immunotherapy plus chemotherapy). FUNDING: Merck Sharp & Dohme, a subsidiary of Merck & Co.