Design: systematic review and random-effects meta-analysis of 22 retrospective studies, PROSPERO registered, with head-to-head and indirect technique comparison
Population: 943 endothelial grafts (656 DSAEK, 287 DMEK) in eyes with prior glaucoma surgery, minimum six months follow-up
Primary outcome: graft survival, with endothelial cell loss, dislocation, rejection and primary failure as secondary outcomes
Effect: 36-month survival 57% (95% CI 48–66) overall and 50% (42–57) with a drainage device; endothelial cell loss 63% at 12 months in those eyes; dislocation 19%; rejection 8–9%; primary failure 4–5%

Endothelial keratoplasty in an eye that has had glaucoma surgery is known to do worse, but the individual series are too small to give a patient a number. A meta-analysis pooled 22 retrospective studies covering 943 grafts — 656 Descemet stripping automated endothelial keratoplasty (DSAEK), 287 Descemet membrane endothelial keratoplasty (DMEK) — with at least six months of follow-up.

The headline figure is sobering. Pooled graft survival at 36 months was 57% (95% confidence interval 48 to 66) across all eyes with prior glaucoma surgery, and 50% (42 to 57) in eyes with a glaucoma drainage device. Endothelial cell loss in drainage-device eyes reached 53% at six months and 63% at twelve. Graft dislocation was more frequent in those eyes (19% against 15% overall). Rejection ran at 8 to 9% and primary graft failure stayed low at 4 to 5%, so the losses are attritional rather than early.

On technique, the analysis is honest about its limits: indirect stratified estimates numerically favoured DSAEK from 24 months, particularly in drainage-device eyes, but head-to-head comparison from studies reporting both showed no significant difference through 24 months. That is not a basis for abandoning DMEK in these eyes. What the figures are a basis for is the consent conversation — a patient with a tube who is being offered a graft should be told that a coin-flip is the realistic three-year expectation, and that a regraft may follow.

Quote 50% three-year survival when consenting an eye with a glaucoma drainage device for endothelial keratoplasty.
Discuss regrafting as a likely part of the plan rather than a complication.
Do not switch technique on this evidence — head-to-head comparison showed no significant difference.
Count and document endothelial cell density at every postoperative visit; the losses are attritional and early trajectory matters.
Consider tube position and its relation to the graft explicitly in surgical planning.

Why it matters

It turns a vague warning that these grafts do badly into a number a surgeon can put in a consent discussion.

The statistics, in plain English

The pooled survival intervals are wide (48–66% overall) because 22 retrospective series with differing follow-up and differing definitions of failure are being combined, and several estimates were reconstructed from published Kaplan-Meier curves rather than taken from raw data. The distinction between the indirect and head-to-head technique comparisons is the important methodological point: indirect comparison across studies is confounded by which surgeons chose which technique in which eyes, while the head-to-head subset — which found nothing — is the more trustworthy analysis and the smaller one.