Identification of the preoperative and perioperative factors that predict postoperative endothelial cell density after Descemet membrane endothelial keratoplasty: A retrospective cohort study.

Low postoperative endothelial-cell density (ECD) plays a key role in graft failure after Descemet-membrane endothelial keratoplasty (DMEK). Identifying pre/perioperative factors that predict postoperative ECD could help improve DMEK outcomes. This retrospective study was conducted with consecutive adult patients with Fuchs-endothelial corneal dystrophy who underwent DMEK in 2015-2019 and were followed for 12 months. Patients underwent concomitant cataract surgery (triple-DMEK) or had previously undergone cataract surgery (pseudophakic-DMEK). Multivariate analyses assessed whether: patient age/sex; graft-donor age; preoperative ECD, mean keratometry, or visual acuity; triple DMEK; surgery duration; surgical difficulties; and need for rebubbling predicted 6- or 12-month ECD in the whole cohort or in subgroups with high/low ECD at 6 or 12 months. The subgroups were generated with the clinically relevant threshold of 1000 cells/mm2. Surgeries were defined as difficult if any part was not standard. In total, 103 eyes (95 patients; average age, 71 years; 62% women) were included. Eighteen eyes involved difficult surgery (14 difficult graft preparation or unfolding cases and four others). Regardless of how the study group was defined, the only pre/perioperative variable that associated significantly with 6- and 12-month ECD was difficult surgery (p = 0.01, 0.02, 0.05, and 0.0009). Difficult surgery also associated with longer surgery duration (p = 0.002). Difficult-surgery subgroup analysis showed that difficult graft dissection associated with lower postoperative ECD (p = 0.03). This association may reflect endothelial cell loss due to excessive graft handling and/or an intrinsic unhealthiness of the endothelial cells in the graft that conferred unwanted physical properties onto the graft that complicated its preparation/unfolding.