Preoperative Evaluation in Descemet Membrane Endothelial Keratoplasty for Secondary Penetrating Keratoplasty Graft Failure

Background: Over the past decade, Penetrating Keratoplasty (PKP) graft failure has been increasingly managed by Descemet Membrane Endothelial Keratoplasty (DMEK). Our aim is to emphasis the importance of preoperative evaluation by Anterior-Segment Optical Coherence Tomography (AS-OCT) and present the clinical outcomes and surgical modications of DMEK performed for Secondary PKP graft failure. Methods: A retrospective medical records review of patients that underwent DMEK for failed PKP at Hadassah Medical Center in 2018-2019. Collected data included demographic characteristics, PKP graft size measured by AS-OCT, corneal donor endothelial cell density (ECD), intra-operative surgical method adjustments, post-operative complications, visual acuity in Snellen (VA), central pachymetry and postoperative ECD. Results: Included were 16 patients (9 males) and 16 eyes. The study period was 18 months. Mean age at performing DMEK was 63 years. Before DMEK, mean VA and central pachymetry were 0.04 and 685µm, respectively. At last follow up, they signicantly improved to 0.3 (p-value=0.001) and 542µm (p-value=0.008) respectively. Mean ECD for donor grafts was 2662 cells per mm 2 . Post-operative ECD was available only for 7 cases with a mean of 1391 cells per mm 2 (p-value=0.0002). At last follow up, 93.75% of the grafts were attached. Graft failure rate was 6.25% due to late decompensation, graft detachment rate and rebubbling rate were 18.75% respectively. Conclusion: A suitable case-based pre-operative evaluation by AS-OCT may play a vital role in DMEK for failed PKP. No less important is to take into consideration multiple surgical adjustments. Both may further decrease complications rates along with accelerating visual recovery. EK: Endothelial keratoplasty (EK); DMEK: Descemet Membrane Endothelial Keratoplasty; DSEK/DSAEK: Descemet stripping (automated) endothelial keratoplasty; PKP/PK: Penetrating Keratoplasty; Re-PKP: Repeating Penetrating Keratoplasty; AS-OCT: Anterior-Segment Optical Coherence Tomography; ECD: endothelial cell density; VA: visual acuity; CME: Cystoid macular edema.

performing DMEK was 63 years. Before DMEK, mean VA and central pachymetry were 0.04 and 685µm, respectively. At last follow up, they signi cantly improved to 0.3 (p-value=0.001) and 542µm (p-value=0.008) respectively. Mean ECD for donor grafts was 2662 cells per mm 2 . Post-operative ECD was available only for 7 cases with a mean of 1391 cells per mm 2 (p-value=0.0002). At last follow up, 93.75% of the grafts were attached. Graft failure rate was 6.25% due to late decompensation, graft detachment rate and rebubbling rate were 18.75% respectively.
Conclusion: A suitable case-based pre-operative evaluation by AS-OCT may play a vital role in DMEK for failed PKP. No less important is to take into consideration multiple surgical adjustments. Both may further decrease complications rates along with accelerating visual recovery.

Background
Endothelial keratoplasty (EK) has recently became a viable surgical treatment in patients with secondary graft failure after penetrating keratoplasty (PK) [1]. About a decade ago, secondary PK graft failure was mostly treated by repeating full-thickness PK. However, repeating PKP (re-PKP) has multiple complications including a higher risk for allograft rejection, high risk of infections from loose sutures and their removal, scarring and thinning of the host cornea from sutures, increased risk of ocular surface disease and slow visual recovery [2,3].
In comparison with DSEK/DSAEK, DMEK provides more selective replacement of the corneal endothelium resulting in smaller incisions, lower risk of allograft rejection and better visual and clinical outcomes [15]. Furthermore, the thinner and more exible DMEK graft may be better suitable for positioning underneath a failed PK graft. It may achieve better apposition than the stiffer DSEK graft, favorable adjustment to the irregular posterior surface or across the posterior PK wound and covers a bigger surface area [16,17]. However, the outcomes of DMEK for failed PK graft may not equal those of primary DMEK. Einan-Lifshitz et al. [18] found higher primary failure rates in DMEK compared with those reported for DSAEK after PKP attributed to persistent postoperative graft detachment. They also found high long-term failure rate, 43% of eyes in the DMEK group and 50% of eyes in the DSAEK group. This may be explained by immunesensitized eyes due to previous transplant, higher rebubbling rates, intraoperative over-manipulation of the DMEK graft due to compromised anterior chamber structures and a different type of wound-healing response [19].
Our study aims to emphasize the importance of preoperative evaluation by Anterior-Segment Optical Coherence Tomography (AS-OCT) and present the clinical outcomes of 16 DMEK surgeries performed for secondary PK graft failure. In addition, the study focuses on adjustments and modi cations to our surgical technique.

Pre-Operative Graft Size Assessment and Donor Tissue Preparation
Each patient underwent Anterior-Segment Optical Coherence Tomography (AS-OCT -Casia 2 Cornea/Anterior Segment OCT and Accessories, TOMEY GmbH) pre-operatively in order to assess the optimal DMEK graft size according to the original PK graft size and morphological features such as bulging posterior graft-host interphase scarring and anterior synechiae, the latter were assessed to consider removing them during the surgery (Figs. 1 and 2).
From donor globes, corneoscleral buttons were excised less than 24 hours postmortem and stored in organ culture medium at 5 °C. Endothelial cell density was checked using Konan CellChek D Donor Cornea Analytics CD-15 specular microscope. The basic standardized donor tissue preparation technique has been already described [19]. DMEK graft sizes ranged between 7.5-8.5 mm diameter. The required DMEK graft size was assessed by AS-OCT (Figs. 1 and 2) according to pre-operatively PK graft size.

Surgical Technique and Multiple Adjustments
The basic standardized DMEK surgical technique has been already described [20]. However, variability in the posterior corneal surface of the recipient due to the presence of the PK graft, as well as the potential restrictions of the PK graft-host junction, required some adjustments and particular manipulations.
Pre-operative sub-tenon Triamcinolone 40 mg was injected at the beginning of the surgery after sub-tenon local anaesthesia. The 2.4-mm wide corneal incision was performed in the host peripheral corneal rim without penetrating the PK graft. Descemetorhexis was started from the center of the PK graft and was completed in a curvilinear pattern along the PK wound in a manner resembling capsulorhexis. It was performed under air or uid maintainer using a reversed Sinskey hook. If scars in the graft-host interphase or anterior synechia existed, they were frequently removed with reversed Sinskey hook, intraocular serrated tweezers, intraocular retinal scissors or with a vitrectomy probe accordingly (Fig. 3, A-C).
The donor Descemet-roll was then injected to the anterior chamber preferably, in a double scroll fashion.
After con rming the correct orientation of the graft by intraoperative AS-OCT (Rescan 700 by Zeiss), when available, or the "Montsouris" sign where the graft is rolled over the tip of the cannula and assures correct orientation of endothelial side down and donor descemet membrane facing the recipient posterior stroma, the graft was unfolded by careful indirect manipulations with air and uid [20]. In most of the more complicated cases, to prevent over-manipulations of the graft, a small air bubble was injected underneath the donor DM to position the tissue onto the recipient posterior stroma and then centering the graft by the "wave maneuver"; gentle tap and horizontal slide on the top of the cornea with the 27G cannula tip in a way like the "L" letter ( Fig. 3, D and E). Some of the grafts were under-sized, in purpose not to be positioned underneath the PK graft-host junction. In cases with a glaucoma drainage device, the presence of the tube in the anterior chamber required speci c additional maneuvers [4].
At the end, the anterior chamber was lled 80-100% with 20% SF6 gas and the eye was left pressurized (Fig. 3F). After 2-3 hours, the patient was checked on slit lamp to evaluate the intraocular pressure. If needed, a release of uid/air/gas from the anterior chamber by pressing on the lower paracentesis with a 27-gauge needle was performed.

Postoperative Management
Postoperative visual acuity was measured using the Snellen visual acuity chart in decimals, postoperative AS-OCT was done to evaluate graft attachment and to measure central pachymetry (Fig. 2). In case of DMEK graft detachment of more than a third of the graft surface area or central detachment affecting the visual axis, a rebubbling procedure was indicated. The main outcome measures were visual acuity, central pachymetry, rebubbling rate and complications after surgery. Endothelial cell density was checked, when the patient could afford performing the test, using CellChek XL Konan Medical specular microscope. Cystoid macular edema (CME) was assessed one month after surgery using Spectralis Spectral-Domain OCT machine (SD-OCT, Heidelberg Engineering, Heidelberg, Germany).

Statistical Analysis
Data were recorded in Microsoft Excel and analyzed using Microsoft Excel and GraphPad Prism 7. For statistical signi cance testing of interval scale parameters, the One-way-ANOVA test was used to compare results preoperatively, one month after surgery and at last follow-up regarding visual acuity and central pachymetry. T-test was used to compare between ECD of the donor and ECD post-operatively. The threshold for statistical signi cance was de ned as P-value < 0.05.

Clinical Outcomes
All surgeries ended with the DMEK graft attached to the posterior part of the PK graft with SF6 bubble supporting the graft. AS-OCT was performed on the same day or a day after the surgery and showed complete attachment of all DMEK grafts.
Mean visual acuity before performing DMEK was 0.04, one month after DMEK it increased to 0.16 (Pvalue = 0.08) and at last follow-up it increased to 0.3 and was statistically signi cant (P-value = 0.001).
Regarding mean central pachymetry, measured by AS-OCT, before performing DMEK it was 685 µm, one month after DMEK it decreased to 574 µm and was statistically signi cant (P-value = 0.04) and at last follow-up it decreased to 542 µm and was statistically signi cant (P-value = 0.008) ( Table 2).  (Table 3). Graft failure was observed only in one eye 10 months after surgery (failure rate was 6.25%). It was diagnosed with late decompensation and treated successfully with a second DMEK. Three eyes had a partially detached graft during their follow up period and underwent a rebubbling procedure (signi cant detachment of more than 1/3 of the graft and rebubbling rates were 18.75% respectively). All rebubbling procedures were successful with fully attached grafts afterward at their last follow-up; at least 4 months after the procedure. In addition, no rejection episodes were observed along the study follow up period, except in one eye that at last two follow up visits, early rejection signs were suspected. Therefore, immediate anti-in ammatory corticosteroids treatment was administered which successfully hindered the immune reaction and kept the cornea clear. In order to evaluate post-operative CME, posterior segment SD-OCT for the macula was performed one month after surgery. CME was observed only in one eye out of 16 (6.25%, case 12).

Discussion
Preoperative evaluation is fundamentally important in any medical procedure [21]. We believe that using AS-OCT has a vital role in adjusting the surgical steps for each case individually in DMEK for PK graft failure.
Common conclusions have emerged from multiple previous studies regarding the results of DMEK after PKP compared to primary DMEK [16,18,19,22,23], including higher rebubbling rates, higher endothelial cell density loss, higher late graft failure rates and more demanding surgical technique. Furthermore, glaucoma ltering surgeries, which are probably more prevalent in post-PKP eyes, are considered a signi cant risk factor for late graft failure, probably as a result of altering the microenvironment in the aqueous humor which accelerates endothelial cell loss [24]. Therefore, based on these eminent conclusions and our experience, additional comprehensive pre-operative AS-OCT evaluation was invested to carefully plan DMEK surgery for each case individually, as well as applying multiple adjustments to our surgical technique, in order to decrease failure rates.
Pre-operative assessment by AS-OCT aimed to evaluate both the PKP graft size and the presence of undesired posterior morphological features, including irregular bulging scars in the graft-host interphase or anterior synechiae (Figs. 1 and 2). Lavy et al. [19] showed that higher detachment rates occurred in oversized grafts. Posterior bulging of graft-host interphase scarring prevents proper graft attachment while anterior synechiae mainly impede intraoperative unfolding DMEK graft and its proper positioning.
Therefore, in cases with posterior bulging and no intraoperative plan to dissect the extra tissue, it is important to plan for undersized DMEK graft to reduce the chances of positioning it underneath the PKhost interphase and enhance DMEK graft attachment post-operatively (Fig. 2, C and D). Moreover, evaluating the presence of posterior morphological features and considering their removal during surgery, may decrease intraoperative graft manipulations and facilitate its positioning (Fig. 2, E and F). In a smooth, non-bulging posterior surface of the corneal graft host interphase, it is worth considering oversizing the DMEK graft, and deliver more endothelial cells to the decompensated cornea (Fig. 2, A and  B).
Intra-operative adjustments included injecting sub-tenon triamcinolone prophylactically (Fig. 3A). We believe that it signi cantly decreases post-operative intraocular in ammation affecting the early postoperative graft failure rate and the incidence of CME. CME is a well-known complication after intraocular surgery and has been reported to occur in 2.0-12.5% of cases after endothelial keratoplasty [25][26][27]. Heinzelmann et al. [28] observed a considerably elevated incidence of CME (13%) which in uence visual rehabilitation. In our study, SD-OCT for macula was performed 1 month after surgery and CME was observed only in one eye (6.25%), which might be explained by the complicated clinical course of this case. Postoperative intraocular in ammation may also affect graft failure and rejection rates which are not negligible after secondary endothelial keratoplasties. Administrating a depo of corticosteroids during the surgery may also have a bene cial effect on graft survival and endothelial cells function.
In addition, paracentesis and main incision were performed in the host peripheral corneal rim without penetrating the PK graft to prevent potential graft-host wound dehiscence (Fig. 3B). No circumferential scoring of DM was performed and descemetorhexis was started from the center of the PK graft and completed in a curvilinear pattern along the PK wound in a manner resembling capsulorhexis (Fig. 3C). We believe, performing these adjustments without removing graft-host interphase sutures, may ensue in a smoother back surface, reduce the amount of Descemet remnants and may even facilitate posterior scar tissue removal. Moreover, scars in the graft-host interphase or anterior synechia, were selectively removed during surgery, if possible, based upon pre-operative AS-OCT (Fig. 3C).
After injecting the DMEK graft, the correct orientation was con rmed by intraoperative AS-OCT (Rescan), when available, or the "Montsouris" sign. Afterward, the graft was unfolded using careful indirect manipulations by tapping on the cornea surface. Then, the graft was elevated with a small air bubble beneath and was centered by the "wave maneuver", an indirect L shaped tapping on the corneal surface (Fig. 3, D and E). We believe that these surgical steps lessen unnecessary extra manipulations on the graft and facilitate positioning it in the correct orientation and suitable position. This way of manipulating the graft on the posterior corneal surface with Descemet-stromal touch, helps to avoid endothelial-iris/IOL touch, which may also protect the graft endothelial cells.
Finally, the anterior chamber was 80-100% lled with 20% SF6 gas to pressurize the eye and support graft adherence for a longer period than air (Fig. 3F). After Two to three hours, intraocular pressure was checked and if necessary, an appropriate intervention was applied. Lavy et al. [19] attributed the tendency of delayed and extensive graft detachment, to insu cient pressurization during surgery or postoperative hypotonia. They recommended to pay attention for pressurizing the eye at the end of surgery or to extend the air-bubble time if enough pressurization cannot be achieved. Compared to other studies, we had relatively low rebubbling rate (18.75%), this may be attributed to SF6 usage and the extended volume in the anterior chamber at the end of the surgery, but also may be due to a more conservative rebubbling policy of the surgeon.
In our study, we encountered relatively low complications rates. Graft failure rate was 6.25% (late), graft detachment rate and rebubbling rate were 18.75% respectively. Interestingly, glaucoma ltering surgeries were not a signi cant risk factor for graft failure. However, this conclusion may be applicable at least regarding early decompensation events and not for late decompensation events due to short-term study period. We believe, based on our experience, that these successful results are attributed to our careful preoperative AS-OCT evaluation and intra-operative adjustments to the surgical technique in DMEK after PKP graft failure. In comparison to our results, higher rebubbling and failure rates were reported by other studies. Lifshitz et al. [18] had 43% rebubbling and 43% failure rates. Lavy et al. [19] found 34% and 36% rebubbling and failure rates respectively. Heinzelmann et al. [23] reported 37% and 21% rebubbling and failure rates respectively.
Visual recovery and central pachymetry improvement were relatively fast during the study period (Table 2). Those favorable clinical outcomes are consistent with other studies as well [16,18,22]. The visual recovery with DMEK under failed PK contrasts sharply with the delayed and unpredictable visual rehabilitation after re-PKP [29]. Moreover, the rapid visual recovery seen after DMEK under failed PK is consistent with the visual recovery that DMEK provides in virgin eyes as compared with PK [30].
There are some limitations to our study, including its retrospective nature, small sample size and shortterm follow up. Moreover, visual acuity tests were done by technicians and not experienced optometrists.
Therefore, uncorrected or partially corrected, rather than best corrected, visual acuity was assessed. In addition, ECD was done only for 7 cases after surgery and comparison between donors ECD and postoperative ECD may be inconclusive due to group size difference. This was mainly because specular microscopy test was not covered by the health insurance and not all study participants could afford it. However, we encountered a predictable decrease in ECD after surgery which may be attributed to cell migration and postoperative in ammatory response [19].

Conclusions
DMEK is a viable method to treat secondary PKP graft failure. It may provide faster and better visual recovery and clinical outcomes compared with re-PKP. Preoperative evaluation using AS-OCT, where available, plays a key role in planning the surgery based on each case characteristics. No less important is adhering to surgical adjustments during DMEK surgery. Both may further decrease the incidence of graft detachment, rebubbling and failure rates.   A -Sub-tenon Triamcinolone 40 mg injected at the beginning of the surgery after sub-tenon local anaesthesia, B -2.4-mm wide corneal incision performed in the host peripheral corneal rim, C -Descemetorhexis performed from the center of the PK graft and completed in a curvilinear pattern along the PK wound, D -Injected donor descemet-roll to the anterior chamber in a double scroll fashion, and con rmed correct orientation by the "Montsouris" sign, E -A small air bubble injected underneath the donor DM and cented graft by "wave maneuver", F -At the end of the surgery, the anterior chamber lled 80-100% with 20% SF6 gas.