Is there evidence for a surgeon learning curve for endothelial keratoplasty in Australia

Importance Expected outcomes from endokeratoplasty may vary with surgeon experience. Background We explored whether a surgeon learning curve exists for Descemet stripping endothelial keratoplasties (manual or automated) performed in Australia. Design Prospective cohort study; various clinical settings. Participants 2139 recipients of 2615 endothelial grafts, registered by 85 surgeons between January 2006 and December 2013. Methods Kaplan-Meier survival analyses and Cox proportional hazards regression were used to examine longitudinal graft survival. Manual and automated Descemet stripping endothelial keratoplasties were analysed together. Pearson Chi-squared analyses were performed to examine differences amongst groups. Continuity correction was used for 2 x 2 tests and statistical significance was set at p < 0.05 (two-sided). Main outcome measure Endothelial graft survival. Results Survival of the first 56 registered grafts was significantly poorer than survival of subsequent grafts (Chi2 = 8.83, df = 1, p = 0.003), when data were combined for all surgeons. Surgeon work-load influenced graft survival significantly (p < 0.001). This variable was retained in multivariate analysis designed to investigate independent factors influencing graft survival. Primary non-functioning grafts were significantly less likely to be reported for endokeratoplasties performed by surgeons with more than 56 registered grafts, compared with those registering 56 or fewer grafts (4.3% versus 8.5%; Chi2 = 18.38, df = 1, p < 0.001). Conclusions and Relevance Our findings suggest that for less experienced or low-volume surgeons, longitudinal graft survival improved once 56 or more endokeratoplasties had been performed, indicative of a learning curve. The learning curve was less apparent for surgeons with 57 or more DSEKs and/or DSAEKs registered during the 8 year study period. Different learning curves may be anticipated for these two groups of surgeons.