Abstract
Purpose :
To identify pre-operative characteristics in penetrating keratoplasty (PK) predictive of graft survival in order to develop a standardized web-based outcome risk calculator.
Methods :
A retrospective chart review was carried out at the Duke Eye Center using procedure code 65730. Inclusion criteria were adults that underwent full-thickness PK from January 1, 2013, till December 31, 2017, that had at least a 4-year follow-up or failed within. Lack of success was concluded if the patient underwent rejection (acute or chronic) despite anti-inflammatory therapy, graft failure, graft dehiscence, visually significant edema, or complications requiring repeat transplantation. Univariable analysis of pre-operative characteristics was performed followed by a stepwise multivariable logistic regression to obtain a predictive model with a p<0.05 cutoff. Discrimination and calibration were assessed using the area under the curve (AUC) and Hosmer-Lemeshow (HL) test, respectively.
Results :
178 transplants were included of which 62 (34.8%) were unsuccessful. History of anterior segment surgery (same eye) (p<0.001), diabetes (p=0.008), smoking (p<0.001), corneal neovascularization (p<0.001), conjunctival injection (p<0.001), synechiae presence (p=0.002), and indication other than keratoconus (p<0.001) were associated with non-success in the univariable analysis. Multivariable analysis produced a model consisting of: (1) previous ocular surface surgery in the same eye (OR: 2.09, p=0.06), (2) diabetes mellitus (OR: 3.48, p=0.013), (3) use of tobacco (OR: 3.77, p=0.007), (4) KNV (i.e. corneal neovascularization) (OR: 2.42, p= 0.042), and (5) erythematous conjunctiva (i.e. conjunctival injection) (OR: 4.36, p=0.001). The predictive model was summarized by the acronym P-DUKE. The AUC achieved was 0.82 and the HL test indicated adequate calibration (p>0.05). The calculator was then deployed under the URL: https://pduke.shinyapps.io/PDUKE/
Conclusions :
The P-DUKE calculator is a readily accessible tool allowing for a standardized assessment of risk in patients planned for PKs and can supplement clinical judgment regarding follow-up and patient expectations irrespective of physician expertise.
This abstract was presented at the 2022 ARVO Annual Meeting, held in Denver, CO, May 1-4, 2022, and virtually.