Descemet membrane endothelial keratoplasty compared with ultrathin Descemet stripping automated endothelial keratoplasty: a meta-analysis

Introduction

Corneal diseases are the third-leading cause of blindness worldwide.1 Fuchs’ endothelial dystrophy (FED), pseudophakic bullous keratopathy (PBK) and failed posterior lamellar keratoplasty are the leading causes of corneal pathologies requiring corneal transplantation in Western countries.2–6 Descemet stripping endothelial keratoplasty (DSEK),7 or its automated cutting version Descemet stripping automated endothelial keratoplasty (DSAEK), and Descemet membrane endothelial keratoplasty (DMEK) are the leading surgical treatments for endothelial failure.8 Compared with DSAEK, DMEK is technically more challenging, has more contraindications,9 higher risk of intraoperative tissue loss4 and might increase the risk of graft dislocation necessitating rebubbling for graft repositioning.2 However, DMEK offers a faster and greater improvement in best-corrected visual acuity (BCVA) and lower rejection rates,1 9 10 likely due to the lack of donor stroma.9 11–13

Thinner DSAEK grafts with less stromal tissue, including ultra-thin-DSAEK (UT-DSAEK), micro-thin-DSAEK and nano-thin-DSAEK were developed to combine the benefits of both procedures.14–17 Studies comparing the results of UT-DSAEK and DMEK showed contradicting results.18–25 These studies differed significantly in methodology, including the study design, inclusion criteria, sample size and DSAEK graft thickness. Recent meta-analyses found that patients who underwent DMEK had a better 12 months BCVA compared with UT-DSAEK, although the rate of complications was higher in the former.26–29 Evidence from meta-analyses is usually ranked at the highest level in clinical practice.30 However, combining data from different studies entails methodological challenges necessitating careful analysis and preplanned protocols.31–33 Among these challenges are the assessment of heterogeneity and exploring its sources, as well as minimising and evaluating publication bias.32–34 Nevertheless, these aspects were not adequately addressed in prior meta-analyses.26–29 Accordingly, gaps in knowledge remain particularly regarding which specific factors are most decisive for the clinical success of endothelial keratoplasty.

The main objective of this meta-analysis was to compare the 12-month logarithm of the minimum angle of resolution (logMAR) BCVA of UT-DSAEK and DMEK for patients with corneal endothelial dysfunction due to FED or PBK. We also explored the differences between the two procedures in the occurrence of complications including graft rejection, graft failure, graft dislocation, rebubbling and glaucoma, and the vision-related quality of life (QOL). We assumed that other recipient or graft characteristics may play a role, such as baseline BCVA, specific UT-DSAEK graft thickness and indication for transplantation, thus conducted subgroup analyses, including a comparison of the outcomes between DSAEK grafts <70 µm and DMEK. Finally, we reviewed all prior meta-analyses compared with our study to synthesise comprehensive evidence on the outcomes of UT-DSAEK and DMEK.

Methods

A systematic literature review and meta-analysis was undertaken following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement.35 The study protocol was registered in PROSPERO.36

Results

The systematic literature search yielded 144 records (online supplemental file 2), of those 8 studies met the inclusion criteria (figure 1), totalling 376 eyes, 189 and 187 in the UT-DSAEK and DMEK groups, respectively. This included three RCTs,19–21 two fellow-eye studies22 23 and three cohort studies.24 25 39 In both fellow-eye studies, UT-DSAEK was performed before DMEK. Four studies were prospective (the three RCTs and one cohort study)19–21 24 (table 1).

Request permission

Methodological issues which might introduce bias were raised during quality assessment (online supplemental file 3). All three RCTs were registered, and study protocols were available for the quality assessment and risk of bias evaluation.40–45 The risk of bias was low in two RCTs,19 21 and in one RCT, there were some concerns regarding the patients selection and the reported result,20 which differed from former publications from this trial43 46 (figure 2A). The risk of bias was serious in one cohort study,22 and critical in four cohort study,23–25 39 mainly due to confounding. Other issues included the lack of assessors blinding and concerns regarding the reported result (figure 2B).

Request permission

Meta-analysis

LogMAR BCVA

The baseline BCVA did not differ significantly between the UT-DSAEK and DMEK groups (combined mean difference −0.01 (95% CI −0.06 to 0.03).19–22 24 25

The 12-month logMAR BCVA was not available in two studies.23 39 Combining data from 6 studies (308 eyes), using the random-effect model, DMEK showed a significantly better BCVA than UT-DSAEK 12-month postsurgery, with a mean difference −0.06 (95% CI −0.10 to –0.02) (figure 3A). Heterogeneity across studies was significant (χ²=11.65, p=0.04) and moderate (I²=57%). Limiting the analysis to prospective data yielded a similar result (mean difference −0.06 (95% CI −0.11 to –0.01) (figure 3B). A sensitivity analysis that included 112 eyes from 2 studies21 24 comparing DMEK only with very thin DSAEK grafts (mean thickness <70 µm) showed no significant difference in 12-month BCVA. The results of BCBA at 6-month postsurgery were consistent with those of the 12 months (online supplemental file 4).

Request permission

Figure 3

LogMAR BCVA at 12 months postsurgery (A) based on all included studies and (B) based only on prospective data (RCTs and prospective cohort studies). BCVA, best-corrected visual acuity; DMEK, Descemet membrane endothelial keratoplasty; IV, inverse variance; LogMAR, logarithm of the minimum angle of resolution; RCT, randomised controlled trial; UT-DSAEK, ultrathin Descemet stripping automated endothelial keratoplast.

ECD

Data regarding 12 months ECD were available from 4 studies (196 eyes). No significant difference was found in ECD between DMEK and UT-DSAEK (mean difference 18.48_cells/mm2 (95% CI −195.99 to 232.95)). Heterogeneity was significant (χ²=14.79, p=0.002) and high (I²=80%). Results were similar when limiting the analysis to prospective data (mean difference −55.18_cells/mm2 (95% CI −342.39 to 232.04)), and consistent for the 6-month postsurgery analysis (online supplemental file 5).

Complications

Data regarding graft rejection 12-month post-transplantation were available for 300 eyes (6 studies).19–22 24 39 Graft rejection was reported in one eye in the UT-DSAEK group.21 The combined risk difference was −0.01 (95% CI −0.04 to 0.03).

There were 2 events of graft failure in the UT-DSAEK group and 6 events (in 5 eyes) in the DMEK group, combining data on 320 eyes (7 studies), yielding a risk difference of 0.02 (95% CI −0.02 to 0.07) and pooled OR=2.32 (95% CI 0.58 to 9.36) favouring UT-DSAEK, when including data from all available studies. The result was similar when analysing only prospective data (figure 4A).

Request permission

Figure 4

Differences in complications incidences at 12 months postsurgery: OR for (A) graft failure (i) based on all included studies and (ii) based on data from prospective studies only (RCTs and prospective cohort studies), (B) graft dislocation (i) based on all included studies and (ii) based on data from prospective studies only and (C) rebubbling (i) based on all included studies and (ii) based on data from prospective studies only*. *Meta-analyses were conducted using the fixed-effects model. DMEK, Descemet membrane endothelial keratoplasty; MH, Mantel-Haenszel; RCT, randomised controlled trial; UT-DSAEK, ultrathin Descemet stripping automated endothelial keratoplasty.

Information on graft dislocation was available on 300 eyes (6 studies).19–22 24 39 There were 14 vs 9 events of graft dislocation in the DMEK group vs the UT-DSAEK group, but the difference was not significant19–21 24 (figure 4B).

Rebubbling risk was calculated using data of 376 eyes (8 studies).19–25 39 Overall 37 (19.79%) and 17 (8.99%) rebubbling procedures were performed in the DMEK and UT-DSAEK groups, respectively. For the risk of need for 1 rebubbling the number needed to harm by choosing DMEK over UT-DSAEK is 9 eyes. The pooled OR for post-DMEK rebubbling was 2.76 (95% CI 1.46 to 5.22), significantly favouring UT-DSAEK, with similar results when including prospective data only19–21 24 (figure 4C).

For glaucoma incidence, combining data from 6 studies,19–22 24 39 showed 4/152 and 4/148 events in the DMEK and UT-DSAEK, respectively (risk difference 0.00 (95% CI −0.04 to 0.04)).

The risk differences results for the complications are shown in online supplemental file 6.

Discussion

The main findings of this meta-analysis show that patients who underwent DMEK had better BCVA 6 months and 12 months after corneal transplantation than patients who had UT-DSAEK. The ECD postsurgery was comparable between the two groups, but the risk of graft dislocation warranting rebubbling was higher in the DMEK group. These results were consistent between the analysis of all included studies and the analysis of only prospective data.

The advantage of DMEK in 12-month BCVA remained stable in most of the subanalyses, except for when we included only cohort studies, or when limiting the analysis to studies with a mean DSAEK graft thickness of <70 µm, in which there was no significant difference between the two procedures. While UT-DSAEK leads to better results than DSAEK,11 our meta-analysis showed that it does not fully compare with the visual results of DMEK. However, the results of the analysis limited to DSAEK grafts of <70 µm thickness may suggest that thinner grafts (nanothin) could potentially fully compare with DMEK. Another meta-analysis compared the results of thin DSAEK graft with <80 µm, 80–100 µm, 100–130 µm thicknesses, showed no difference in the clinical outcome including BCVA between the groups.48

As for the cohort studies-only analysis, these studies have the potential for better generalisability only if they have a robust methodology, as they represent real-world data. Nonetheless, a major limitation of observational studies is the lack of random allocation and confounding. Indeed, the cohort studies included in our meta-analysis were classified as having serious or critical risk of bias,22–25 39 thus limiting their generalisability.

We found no difference in ECD between the procedures. At 6 months, the study by Chamberlain et al19 introduced heterogeneity, probably because of its result rather than its methodology.

Graft dislocation is common following both DMEK and UT-DSAEK, and often requires further intervention such as rebubbling.49–51 Complete detachment sometimes necessitates a repeat transplantation.50 51 We found a higher rate of graft dislocation after DMEK, but the difference from UT-DSAEK was not significant. Notably, some DMEK dislocations, including peripheral dislocations or small dislocations of <1/3 of the graft surface area, often resolve spontaneously.50 52–54 Nevertheless, we found a significantly higher risk of post-DMEK rebubbling than after UT-DSAEK. In a study that used machine learning approaches to recognise risk factors for graft detachment based on all (n=3647) the posterior lamellar keratoplasties recorded in the Dutch Cornea Transplant Registry between 2015 and 2018, DMEK was found to be a risk factor for graft detachment.49 In the results from the DMEK report based on the Netherlands Organ Transplant Registry, rebubbling rate was 19%,55 similar to the 19.79% observed in our meta-analysis. It is unclear whether rebubbling decreases ECD or reduces the graft survival, as results were contradicting.54 56–59 The increased rebubbling risk after DMEK should be considered, and close observation after surgery is needed.

In the current meta-analysis, graft rejection was rare, and graft failure was also low, possibly due to the short follow-up period (12 months), thus a longer follow-up period is needed to better understand the difference between the procedures in these outcomes. Three recent large studies using real-world data reported worse survival rates of DMEK grafts vs DSAEK (or DSEK).6 8 55 The study using the Australian Corneal Graft Registry had the longest follow-up, and the advantage of DSAEK was consistent in all time points (1, 2, 4 and 6 years post-transplantation).8 All these studies suggested that the ongoing learning curve could be a possible reason for the worse survival of DMEK grafts,6 8 55 as DMEK was adopted more recently than DSAEK.8 55 However, a retrospective cohort study comparing the 5-year survival of DMEK, DSAEK and penetrating keratoplasty, showed superior survival for DMEK over DSAEK grafts throughout the follow-up.60 Yet, this finding was prone to confounding due to varying predominant indications for surgery: DMEK was performed mainly on FED cases, while DSAEK mostly addressed PBK cases, which showed lower survival rates. They also reported that in eyes with PBK, DMEK had better survival and survival remaind stable after the first year, unlike declining DSAEK graft survival, suggesting that failure of DMEK after the first year is rare. Yet, this finding was limited by a very small number of PBK cases in the DMEK group. A subgroup analysis showed no difference between UT-DSAEK and DMEK grafts. Price et al in their retrospective study reported similar 5-year graft survival rates for DSEK and DMEK, although DMEK exhibited a significantly lower rejection risk, and most rejection episodes responded well to topical corticosteroids.61 It is possible that long-term data from larger cohorts on DMEK grafts transplanted after mastering this technique, may show an advantage over DSAEK. More evidence on long-term survival of UT-DSAEK versus DMEK are needed to better understand the potential adverse events of these procedures.

Three of the included studies reported vision-related QOL, only one of them reported a significant difference between DMEK and UT-DSAEK, with a faster subjective recovery and an overall preference for DMEK.22 QOL was also reported separately for two other studies inculded in the current review.62 63 Nevertheless, varying questionnaires were used,21–23 62 63 limiting the ability to integrate the results.

Collectively, this evidence emphasises that the decision regrading the choice of surgical method for endothelial corneal transplantation remains multifaceted. After considering the indication for keratoplasty, ocular comorbidities and surgical expirience with the different procedures, patients should be informed regarding the expected visual results and risk of complication, and understand the expected postoperative course. For example, need for frequent controls for early diagnosis of graft dislocations and need for rebubbling, or a faster expected recovery and better vision-related QOL. These factors should be carefully discussed to achieve an optimal shared decision-making and improve surgical outcomes and patient satisfaction.

Our results are in line with prior meta-analyses comparing DMEK and UT-DSAEK.26–29 Our study has several strengths that distinguish it from the others conducted on this topic,26–29 including the prospective registration in PROSPERO, enabling comparison with the predesigned study protocol and enhancing internal validity, the use of the most recommended means for evaluating risk of bias (ROB 2 for RCTs and ROBINS-I for cohort studies), and a thorough sensitivity analysis which allowed us to identify sources of heterogeneity between the studies, and to validate our results. Additionally, the methodology and inclusion criteria differed across the previous meta-analyses. Maier et al26 did not include the study by Machalińska et al,39 due to missing data on the outcomes and baseline differences in BCVA between groups. We included this study on the complications rate. Dimtsas et al27 excluded from their meta-analysis the studies by Torras-Sanvicens et al23 and Kurji et al,24 and reported they found the latter ineligible as they considered nanothin DSAEK as a different entity from UT-DSAEK. This approach is challenged given that the paramount question is whether thinner DSAEK grafts compare with DMEK, and even more as the definitions of UT-DSAEK, microthin-DSAEK and nanothin-DSAEK are inconsistent throughout the literature.9 64 We therefore found the inclusion of the study by Kurji et al24 of value and addressed the thickness differences in the sensitivity analysis, allowing a better understanding of the differences between these procedures. Hurley et al28 did not include the studies by Machalińska et al39 and Kurji et al24 but included the study by Tourabaly et al,65 which did not meet the inclusion criteria of 12 months results in our study. Singh et al conducted their literature search in June 2021,29 prior to the publication of some of the studies included in our review.21 39 Moreover, in the previous meta-analyses, publication bias was not adequately addressed.26–29 The methodological issues of the previous meta-analyses,26–29 and foremost the absence of sensitivity analysis to account for inherent risks when pooling data from heterogenous studies, challenge their findings.

Our study has limitations. Significant heterogeneity was noted across the studies in some analyses. Additionally, the study by Romano et al25 was characterised by a worse mean BCVA, raising a concern regarding its effect on the results. These issues were addressed in the sensitivity analysis. Furthermore, the definition of UT-DSAEK was inconsistent between studies. For example, Dunker et al20 defined UT-DSAEK as targeted central residual graft thickness of 100±20 µm, Matsou et al21 declared achieving a thickness of <130 µm in 100% of grafts, and Chamberlain et al19 defined UT-DSAEK as grafts that were cut to between 60 µm and 90 µm. Namely, not only the values used are inconsistent, but also whether these represent a targeted mean thickness, an upper value, or a range of thicknesses for grafts that are achieved using a specific cutting method. We believe that a standardisation of terminology is warranted for future studies, and suggest, based on our systematic review, that UT-DSAEK should refer to grafts <130 µm and the term nanothin-DSAEK to grafts <50 µm. Moreover, the duration of follow-up in most of the studies included in this review was insufficient to assess the rate of long-term complications. Currently data regarding the long-term survival UT-DSAEK and DMEK is scarce. Publication bias is a generic limitation of any meta-analysis. Since data on the main outcome of BCVA at 12 months was not available for all studies, we also assessed publication bias based on the rebubbling rate data. The significant publication bias for the 12-month BCVA results suggested potential missing studies favouring UT-DSAEK. Nevertheless, for the outcome of rebubbling no publication bias was evident. Two studies reported the rebubbling rate but not the 12-month logMAR BCVA,23 39 one of them showed a tendency towards UT-DSAEK at 6 months.23 This suggests minimal publication bias.

In conclusion, DMEK resulted in a better visual acuity than UT-DSAEK, but with a higher risk of rebubbling, which was required in one-fifth of the patients. This should be considered prior to surgery since DMEK patients should remain under close observation in the postoperative period. Larger studies with longer follow-up are needed to compare the long-term graft survival and the risk of rarer complications such as graft rejection, graft failure or glaucoma. Even thinner DSAEK grafts (mean thickness <70 µm) could potentially compare with DMEK. Standardisation of the nomenclature of thinner DSAEK grafts is warranted.