HIGHLIGHTS
- •
PK may provide better graft survival compared to DSEK and DMEK.
- •
Visual acuity improvements were comparable after PK, DSEK and DMEK.
- •
Glaucoma surgery performed appear to have no significant impact on graft survival.
Purpose
To compare surgical outcomes following penetrating keratoplasty (PK), Descemet stripping endothelial keratoplasty (DSEK), and Descemet membrane endothelial keratoplasty (DMEK) in patients with iridocorneal endothelial (ICE) syndrome.
Design
Systematic review and meta-analysis on individual patient data (IPD).
Methods
Pre-registration was performed in the PROSPERO database (registration number: CRD42024539444). Eligible studies from Embase, MEDLINE (via PubMed), and the Cochrane Central Register of Controlled Trials (CENTRAL) were retrieved up to April 24, 2024. Studies were included those reporting clinical outcomes after PK, DSEK, or DMEK- graft survival, best spectacle-corrected visual acuity (BSCVA) and endothelial cell density (ECD) – in people with ICE syndrome. Cochrane Handbook was followed for data extraction/ synthesis, and the Risk of Bias in Non-randomized Studies of Interventions (ROBINS-I) and the Joanna Briggs Institute Critical Appraisal Checklists were used to assess risk of bias. Meta-analyses were conducted using a random-effects model. Heterogeneity between studies was assessed using Q-test and I 2 statistics.
Results
Nineteen of the 1963 screened studies were included in the meta-analysis. Multivariate pooled Kaplan-Meier curves with 95% confidence intervals, based on IPD from studies with at least 10 cases indicated that graft survival was better after PK compared to DSEK in patients with ICE syndrome. No significant difference ( P = . 92) was found in BSCVA improvement between PK [-0.77 (95% CI, -1.45 to -0.09)], DSEK [-0.87 (95% CI, -1.35 to -0.39)] and DMEK [-0.85 (95% CI, -1.07 to -0.62)]. No significant differences in ECD were observed between DSEK and DMEK 6 ( P = . 88) and 12 months ( P = . 33) postoperatively. IPD analysis revealed no significant difference in graft survival between patients with and without anytime glaucoma (-0.04 ± 0.50 SEM; P = . 940) or cataract surgery (-0.45 ± 0.40 SEM; P = . 265).
Conclusions
PK demonstrated better graft survival compared to DSEK in patients with ICE, however, further research and additional evidence are needed to draw more definitive conclusions. Improvements in BSCVA were comparable across PK, DSEK and DMEK. Glaucoma surgery, whether performed before or after keratoplasty, appear to have no significant impact on graft survival.
INTRODUCTION
I ridocorneal endothelial (ICE) syndrome, first described by Yanoff in 1979, is a rare ocular disorder with 3 distinct clinical subtypes: progressive iris atrophy, Chandler syndrome, and Cogan-Reese syndrome. , Although its cause remains unknown, ICE syndrome is characterized by abnormal corneal endothelium, iris atrophy with nodules, and peripheral anterior synechiae. This unilateral, non-hereditary condition predominantly affects Caucasian women.
In ICE syndrome, the endothelium exhibits a distinctive hammered-silver appearance, which can result in irreversible corneal oedema and decompensation. Pathologic endothelial-like cells, known as ICE cells, proliferate and migrate to the trabecular meshwork, potentially causing glaucoma and vision loss. Corneal oedema in individuals with ICE syndrome is thought to arise from both the impaired pump function of the ICE cells and elevated intraocular pressure. As a result, the management of corneal decompensation in ICE syndrome presents a significant surgical challenge.
Management of ICE syndrome primarily involves addressing corneal oedema and glaucoma. Traditionally, penetrating keratoplasty (PK) was the primary surgical approach for resolving corneal edema and restoring vision. However, PK has shown high rates of graft rejection and failure in ICE syndrome. Endothelial keratoplasty techniques such as Descemet stripping endothelial keratoplasty (DSEK) and Descemet membrane endothelial keratoplasty (DMEK), selectively replace the posterior corneal layers, while preserving the stroma and the epithelium. These approaches offer significant advantages over PK for treating corneal endothelial failure. DSEK and DMEK are considered less invasive than PK, providing better and faster visual recovery. DMEK allows for the selective replacement of the endothelium and Descemet membrane without involving additional corneal stroma, yielding even better surgical outcomes compared to DSEK. However, the limited number of studies and small sample sizes make it challenging to determine the long-term outcomes and graft survival rates for these techniques in ICE syndrome.
This study aims to perform a systematic review and meta-analysis to compare graft survival, visual acuity, and endothelial cell density following PK, DSEK, and DMEK in patients with ICE syndrome. The hypothesis is that significant differences in postoperative outcomes among the 3 keratoplasty techniques may exist, providing valuable insights to guide the optimal surgical approach for managing ICE syndrome.
METHODS
The systematic review and meta-analysis was conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 statement, and was pre-registered in the PROSPERO database (registration number: CRD42024539444). The study methodology adhered to the recommendations outlined in the Cochrane Handbook.
Information Source and Search Strategy
The electronic databases of Embase, MEDLINE (via PubMed), and the Cochrane Central Register of Controlled Trials (CENTRAL) were systematically searched for relevant studies up to April 24, 2024, without language restrictions. The search strategy included the following keywords: (irido-corneal endothelial syndrome) OR (iridocorneal-endothelial syndrome) OR (iridocorneal endothelial syndrome) OR (essential iris atroph*) OR (Chandler’s syndrome) OR (Chandler syndrome) OR (iris nevus syndrome) OR (Cogan-Reese syndrome). No search filters were applied.
Eligibility Criteria
The research question was developed using the Population-Intervention-Comparator-Outcomes (PICO) framework. Eligible studies included patients with ICE syndrome (P) who underwent PK, DSEK/ DS(A)EK, or DMEK (I and C). These studies reported outcomes (O) such as graft survival, pre- and postoperative best-corrected visual acuity, endothelial cell density (ECD) and endothelial cell loss (ECL).
Studies were included if they provided data on at least 1 treatment modality for ICE syndrome and reported the specified outcomes. No distinction was made between DSEK and Descemet stripping automated endothelial keratoplasty (DSAEK) in the included studies. Both retrospective and prospective studies, as well as case series, were eligible for inclusion, while duplicates, case reports, and non-human studies were excluded.
Selection Process
The articles were managed using EndNote 20 reference manager (Clarivate Analytics, Philadelphia, PA, USA). After the automatic and manual removal of duplicates, the titles, abstracts, and full texts were independently screened by 2 authors working in pairs (GT and KK). Any disagreements were resolved by a third reviewer (NS). The full texts of potentially eligible publications were further evaluated for inclusion. In cases of overlapping study populations, the publication with the larger sample size was selected for inclusion.
Data Collection Process and Items
Data from articles meeting the inclusion criteria were extracted into an Excel spreadsheet (Office 2016, Microsoft, Redmond, WA, USA). The extracted data included general information such as author, year, study design, patient demographics (age, gender), number of subjects, and follow-up duration. For each study, baseline (preoperative) and postoperative values were recorded, and outcomes were collected at multiple time points when available.
Extracted data included graft survival, mean best spectacle-corrected visual acuity (BSCVA) converted into logMAR, ECD in cells/mm 2 and ECL in %. Data about simultaneous cataract surgery was also collected.
For outcomes such as graft survival presented in Kaplan-Meier (KM) curves as figures, quantitative data were extracted using the WebPlotDigitizer software ( https://automeris.io/WebPlotDigitizer.html , accessed on 9 November 2024) to enable further analysis.
Where available, individual patient data (IPD) were collected as provided by study authors. Any disagreements in data extraction were discussed and resolved through team consensus.
Quality Assessment of Included Studies
The risk of bias was assessed using the Risk of Bias in Non-randomized Studies of Interventions (ROBINS-I) tool, for non-randomized interventional studies and the Joanna Briggs Institute Critical Appraisal Checklists, for observational studies, case series, and case reports. Disagreements among review authors regarding bias assessments were resolved through discussion. If consensus could not be achieved, a third reviewer was consulted to make the final decision.
Statistical Methods and Data Synthesis
Statistical analyses were carried out using packages “IPDfromKM,” “meta,” “metafor,” “nlme” and “survival” of the R statistical software (version 4.1.2.). The statistical analyses follow the advice of Harrer et al. For all statistical analyses, a p-value of less than 0.05 was considered significant. All the performed analyses included random effect terms.
We performed individual meta-analyses on IPD. We used linear mixed effect regression for this purpose except for the pooled hazard ratio, for which we used the mixed effects (frailty) Cox Proportional Hazards model. We analyzed the impact of the different covariates by including them in the regressions.
For each outcome, only 1 part of the studies reported IPD. Hence, we also performed study-level meta-analyses, including studies reporting IPD data and studies reporting only summary statistics. Classical inverse variance random-effect meta-analysis was applied using the REML tau estimator and Hartung-Knapp adjustment to pool mean and mean differences. We applied this approach to the log-transformed probabilities to pool the 36-month survival (no graft failure) probabilities within a slightly more complex framework. When a study provides graft survival results in 2 subgroups, even though the underlying patients are different, the random effects are correlated. For this reason, we used a 3-level meta-analysis model.
We visualized the pooled outcomes and their 95% confidence and prediction intervals on forest plots. Besides the prediction interval, heterogeneity was assessed by calculating I² measure and its confidence interval and performing the Cochrane Q test. We also performed several study-level subgroup and meta-regression analyses. We visualized meta-regression results on bubble plots. When the meta-regression is based on aggregated variables, then the results must be interpreted with caution due to the possibility of aggregation bias, see section 7.6.2. of Schmid et al. for the details.
We used the WebPlotDigitizer tool to read digitized KM curves. We estimated IPD data from the digitalized KM curves by direct calculation via inverting the construction of the KM curve similarly as described in Guyot et al. We plotted the study-level KM curves on the sample plot. Moreover, we also created pooled survival curves with confidence intervals using the multivariate methodology of Combescure et al.
In the majority of the meta-analyses, only studies with at least 3 observations were included. In case of the study-level meta-analysis of the 36-month survival probability, we only included studies having at least 10 patients. For the individual Cox regression we performed the analyses under both criteria.
A leave-one-out sensitivity analysis was conducted by systematically excluding 1 study at a time to evaluate the robustness of our findings. For the analysis of mean differences in BSCVA, the univariate nature of the data allowed for the generation of leave-one-out plots using the dmetar R package. For all other outcomes, leave-one-out results were presented in tabular format.
Publication bias analysis was not possible since, in each subgroup, the number of involved studies was less than 10.
Assessment of the Grade of Evidence
Given the limited number of comparative studies, it was not possible to determine the level of evidence for the analyzed outcomes.
Protocol Amendment
As there were not enough articles and reported data on rates of rejection and rebubbling, we deviated from the original plan and did not analyse their incidences.
RESULTS
Search and Selection
Our systematic search identified a total of 2813 articles. After removing duplicates, 1963 publications were screened. Ultimately, 19 studies were deemed eligible for inclusion in the qualitative and quantitative synthesis ( Figure 1 ).
Basic characteristics of included studies
Table 1 shows the baseline data of the included studies. We included 1 prospective and 2 retrospective cohort studies involving 311 patients with ICE, along with 16 case series comprising 208 patients. In total, 519 patients were included in the meta-analysis and systematic review. Additionally, we utilized IPD from 13 studies.
| First author (year) | Design | Country | Recruitment period | No. (woman %) | Age (years) | GR 1 | GR 2 | Mean follow up (months) | Outcome |
|---|---|---|---|---|---|---|---|---|---|
| Alvim et al. (2001) | RCS | USA | 1985-1999 | 14 (57.14) | 53; NR | PK | – | 58.21 | GS, BSCVA |
| Ao et al. (2017) | RCS | China | 2008-2015 | 18 (72.2) | 51.1 ± 13.0 | DSEK | – | 19.0 | GS, BSCVA, EDC, ECL |
| Chang et al. (1993) | RCS | USA | 1980-1991 | 12 (66.7) | 52.9 ± 12.9 | PK | – | 30.3 | GS, BSCVA, ECL |
| Chaurasia et al. (2013) | RCS | India | 2009-2011 | 7 (42.9) | 50.4 ± 7.5 | DSEK | – | 12.5 | GS, BSCVA |
| Chaurasia et al. (2021) | RCS | India | NR | 3 (100) | 35.3 ± 6.7 | DSEK | – | 53.3 | GS, BSCVA, EDC |
| Crawford et al. (1989) | RCS | USA | 1975-1983 | 9 (88.9) | 54.6; NR | PK | – | 42 | GS, BSCVA |
| DeBroff et al. (1994) | RCS | USA | 1971-1992 | 6 (50.0) | 64.3 ± 10.7 | PK | – | 45.5 | GS, BSCVA |
| Fajgenbaum et al. (2015) | RCS | United Kingdom | 2006-2014 | 4 (75.0) | 56.8 ± 17.0 | DSEK | – | 55.3 | GS |
| Joshi et al. (2022) | RCS | India | NR | 5 (40.0) | 48.2 ± 10.8 | DMEK | – | 29.6 | GS, BSCVA, EDC, ECL |
| Li et al. (2023) | RCS | China | 2018-2021 | 24 (54.2) | 53.9; NR | DSEK | – | 12.0 | GS, BSCVA, EDC, ECL |
| Mohamed et al. (2022) | RCS | India | 2010-2019 | 52 (63.5) | 48.8 ± 10.8 | DSEK | – | 28.8 | GS, BSCVA, EDC, ECL |
| Price et al. (2007) | RCS | USA | 2005-2006 | 3 (0) | 55.0 ± 10.6 | DSEK | – | 8.3 | GS, BSCVA |
| Quek et al. (2015) | Observational, retrospective cohort study | Singapore; USA | 1991-2011 | 29 (58.6) | 55.3 ± 10.2 | PK | DSEK | 5.2 | GS |
| Roberts et al. (2023) | Observational, prospective, multicenter, cohort study | Australia | 1985-2020 | 196 (58.7) | 56 ± 14 | PK | DSEK, DMEK | NR | GS |
| Rotenberg et al. (2020) | Observational retrospective, multicenter, cohort study | United Kingdom | 2000-2017 | 86 (48.3) | 56.2; NR | PK | DSEK | NR | GS |
| Siddharthan et al. (2020) | RCS | India | NR | 4 (75.0) | 48.8 ± 7.5 | DMEK | – | 36.0 | GS, BSCVA, EDC, ECL |
| Wu et al. (2021) | RCS | China | 2014-2018 | 24 (58.3) | 48.5 ± 6.4 | DMEK | – | 24.9 | GS, BSCVA, EDC, ECL |
| Ziaei et al. (2018) | RCS | New Zealand | NR | 3 (66.6) | 45.7 ± 15.0 | DMEK | . | 12.0 | GS, BSCVA |
| Zhang et al. (2023) | RCS | China | 2015-2022 | 20 (50.0) | 52.5 ± 10.9 | DSEK | – | 18.8 | GS, BSCVA, EDC, ECL |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
