Establishing a Biomarker for the Prediction of Short-Term Graft Detachment After Descemet Membrane Endothelial Keratoplasty

Purpose: The purpose of this study was to investigate the predictive value of pachymetry mapping 1 day after Descemet membrane endothelial keratoplasty (DMEK) as a biomarker for early graft detachment. Methods: This was a post hoc analysis of 65 pseudophakic subjects with Fuchs endothelial dystrophy who underwent DMEK surgery between December 2018 and April 2021 as part of the Advanced Visualization In Corneal Surgery Evaluation international multicenter randomized controlled trial. One eye per patient was included. Preoperatively and 1 day postoperatively, patients underwent anterior segment optical coherence tomography imaging. Using a grid consisting of 25 zones (ie, pachymetry map), corneal thickness and presence of a graft detachment were mapped for each patient. Detachments of any size were considered, regardless of subsequent clinical interventions. Missing data were imputed and subsequently divided into a training and test set. Two prediction methods were evaluated: one model based on absolute corneal thickness and a regression model. Results: A total of 65 eyes were included for analysis of which 33 developed any form of graft detachment. Preoperatively, no significant differences were observed between the groups (P = 0.221). Corneal thickness in the corneal zones with a detached graft was significantly increased compared with corneal zones with an attached graft (P, 0.001). The regression prediction model had an area under the curve of 0.87 (sensitivity: 0.79 and specificity: 0.75), whereas the absolute thickness cutoff model only reached 0.65. Conclusions: Pachymetry mapping 1 day after DMEK was predictive for early graft detachment, and the prediction model had a good to excellent performance. This aids in identifying patients at risk for graft detachment and subsequent tailored postoperative care.