BACKGROUND: Accurate identification of sentinel lymph nodes in patients with cancer improves detection of metastatic disease and decreases surgical morbidity. We sought to establish whether indocyanine green fluorescent dye is non-inferior to isosulfan blue dye in detecting sentinel lymph nodes in women with cervical and uterine cancers. METHODS: In this non-inferiority, within-patient comparison study, patients aged 18 years or older with clinical stage I endometrial or cervical cancer undergoing curative surgery were randomly assigned 1:1 to lymphatic mapping with isosulfan blue dye (visualised by white light) followed by indocyanine green (visualised by near-infrared imaging), or indocyanine green followed by isosulfan blue dye. Permuted block randomisation with stratification by study site was done with a computerised random number generator. All participants were masked to their randomisation assignment until after the procedure; however, investigators were not masked to the procedure used. Laparoscopic surgery with the PINPOINT near-infrared fluorescence imaging system (Stryker, Kalamazoo, MI, USA) was used in all cases. The primary outcome was efficacy of intraoperative indocyanine green with near-infrared fluorescence imaging versus that of isosulfan blue dye in the identification of lymph nodes, defined as the number of lymph nodes identified by indocyanine green and isosulfan blue dye, respectively (and confirmed as lymphoid tissue by histology), divided by the number of lymph nodes identified intraoperatively and excised. The study had a 5% non-inferiority margin needed to show non-inferiority of the frequency of lymph node detection with indocyanine green to that with isosulfan blue dye with 80% power at a 5% two-sided significance level. Analyses were done in both per-protocol and modified intention-to-treat populations. The trial was registered with ClinicalTrials.gov, number NCT02209532, and is completed and closed. FINDINGS: Between Dec 21, 2015, and June 19, 2017, 180 patients were enrolled and randomly assigned to the two groups (90 to each group); 176 patients received the intervention and were evaluable (modified intention-to-treat population). 13 patients with major protocol violations were subsequently excluded from the per-protocol population. 517 sentinel nodes were identified in the per-protocol population (n=163), of which 478 (92%) were confirmed to be lymph nodes on pathological processing: 219 (92%) of 238 nodes that were both blue and green, all seven nodes that were blue only, and 252 (95%) of 265 nodes that were green only (p=0·33). Seven sentinel lymph nodes were neither blue nor green but were removed for appearing suspicious or enlarged on visual examination. In total, 471 (97%) of 485 lymph nodes were identified with the green dye and 226 (47%) with the blue dye (difference 50%, 95% CI 39-62; p<0·0001). In the modified intention-to-treat population (n=176), 545 nodes were identified, of which 513 (94%) were confirmed to be lymph nodes on pathological processing: 229 (92%) of 248 nodes that were both blue and green, all nine nodes that were blue only, and 266 (95%) of 279 nodes that were green only (p=0·30). Nine sentinal lymph nodes were neither blue nor green but were removed for appearing suspicious or enlarged on visual examination. 495 (96%) of 513 nodes were identified with the green dye and 238 (46%) with the blue dye (50%, 39-61; p<0·0001). INTERPRETATION: Indocyanine green dye with near-infrared fluorescence imaging identified more sentinel nodes than isosulfan blue dye in women with cervical and uterine cancers, with no difference in the pathological confirmation of nodal tissue between the two mapping substances. FUNDING: Novadaq.